1 line
89 KiB
JSON
1 line
89 KiB
JSON
{"repo": "assassin187/researchclaw", "files": {"README.md": {"7952e42eaa02bb270e41835b778ef52e1a7ddd93": {"t": 1785747202.8138788, "n": 6550, "h": "62cfbe62de2572b6", "c": "# 🐈 ReasearchClaw\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **AI Paper Review** — Peer-review any PDF against NeurIPS / ICLR / ICML / EuroSys / CVPR rubrics; produces narrative feedback + dimensional scores + Accept/Reject\n- [x] **Paper Discovery** — Infer search keywords from a research direction, search arXiv + Semantic Scholar, and generate Chinese reports for influential and recent papers\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter 3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.reasearchclaw/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n### AI Paper Review\n\nReview any paper PDF against real conference rubrics — produces a structured narrative review (Synopsis, Strengths, Weaknesses, Suggestions, References) plus venue-specific dimensional scores and an Accept/Reject recommendation:\n\n```bash\nexamples/ai_review_command.sh\n```\n\nSupported venues: NeurIPS · ICLR · ICML · EuroSys · OSDI · SOSP · CVPR · ICCV · General\n\n**Accuracy check — ICLR 2024 papers:**\n\n| Paper | Real outcome | Originality | Significance | Contribution | Overall | Decision |\n|-------|-------------|-------------|--------------|--------------|---------|----------|\n| [Mamba](https://arxiv.org/abs/2312.00752) (arXiv:2312.00752) | Accepted (Spotlight) | **4 / 4** | **4 / 4** | **4 / 4** | **8 / 10** | **Accept** |\n| [SELF-RAG](https://arxiv.org/abs/2310.11511) (arXiv:2310.11511) | Rejected from ICLR ¹ | 3 / 4 | 3 / 4 | 3 / 4 | 7 / 10 | Accept |\n\n> ¹ **Note on SELF-RAG:** This is a high-quality paper — it was accepted at EMNLP 2023 and has thousands of citations. Its ICLR 2024 rejection was due to a venue policy (already published elsewhere), **not** a reflection of paper quality. Our system correctly evaluates it as solid, well-executed research (7/10), just not at the same level of architectural breakthrough as Mamba (8/10, with three dimensions at 4/4).\n\nThe accepted paper (Mamba) scores higher on every breakthrough dimension. All three dimensions where Mamba achieves 4/4 — Originality, Significance, Contribution — are the strongest predictors of lasting architectural influence. SELF-RAG scores a uniform 3/4, reflecting solid incremental work. Full review texts: [Mamba](examples/reviews/mamba_iclr2024_accepted.md) · [SELF-RAG](examples/reviews/selfrag_iclr2024_rejected.md).\n\nOutput folder:\n\n```text\n~/.reasearchclaw/workspace/outputs/paper_review/<pdf_stem>/\n├── review.md # 6-section narrative + score table\n└── meta.json # venue, mode, scores, elapsed time\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install reasearchclaw-ai\n```\n\n### 2. Initialize\n\n```bash\nreasearchclaw onboard\n```\n\nThis creates `~/.reasearchclaw/config.json` and a default workspace at `~/.reasearchclaw/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.reasearchclaw/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nreasearchclaw agent -m \"Hello!\"\n\n# Interactive REPL\nreasearchclaw agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `reasearchclaw onboard` | Initialize config and workspace |\n| `reasearchclaw agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `reasearchclaw gateway [-p PORT]` | Start the multi-channel gateway |\n| `reasearchclaw status` | Show config, API keys, and workspace status |\n| `reasearchclaw channels status` | Show channel connection status |\n| `reasearchclaw channels login` | Link WhatsApp via QR code |\n| `reasearchclaw cron list` | List scheduled jobs |\n| `reasearchclaw cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `reasearchclaw cron remove <ID>` | Remove a scheduled job |\n| `reasearchclaw cron enable <ID>` | Enable / `--disable` a job |\n| `reasearchclaw cron run <ID>` | Manually trigger a job |\n"}, "042aee679456628508c6ffb3cb3ca6e8aed5c21e": {"t": 1785771354.1728969, "n": 6424, "h": "be660832489b36bb", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **AI Paper Review** — Peer-review any PDF against NeurIPS / ICLR / ICML / EuroSys / CVPR rubrics; produces narrative feedback + dimensional scores + Accept/Reject\n- [x] **Paper Discovery** — Infer search keywords from a research direction, search arXiv + Semantic Scholar, and generate Chinese reports for influential and recent papers\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter 3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n### AI Paper Review\n\nReview any paper PDF against real conference rubrics — produces a structured narrative review (Synopsis, Strengths, Weaknesses, Suggestions, References) plus venue-specific dimensional scores and an Accept/Reject recommendation:\n\n```bash\nexamples/ai_review_command.sh\n```\n\nSupported venues: NeurIPS · ICLR · ICML · EuroSys · OSDI · SOSP · CVPR · ICCV · General\n\n**Accuracy check — ICLR 2024 papers:**\n\n| Paper | Real outcome | Originality | Significance | Contribution | Overall | Decision |\n|-------|-------------|-------------|--------------|--------------|---------|----------|\n| [Mamba](https://arxiv.org/abs/2312.00752) (arXiv:2312.00752) | Accepted (Spotlight) | **4 / 4** | **4 / 4** | **4 / 4** | **8 / 10** | **Accept** |\n| [SELF-RAG](https://arxiv.org/abs/2310.11511) (arXiv:2310.11511) | Rejected from ICLR ¹ | 3 / 4 | 3 / 4 | 3 / 4 | 7 / 10 | Accept |\n\n> ¹ **Note on SELF-RAG:** This is a high-quality paper — it was accepted at EMNLP 2023 and has thousands of citations. Its ICLR 2024 rejection was due to a venue policy (already published elsewhere), **not** a reflection of paper quality. Our system correctly evaluates it as solid, well-executed research (7/10), just not at the same level of architectural breakthrough as Mamba (8/10, with three dimensions at 4/4).\n\nThe accepted paper (Mamba) scores higher on every breakthrough dimension. All three dimensions where Mamba achieves 4/4 — Originality, Significance, Contribution — are the strongest predictors of lasting architectural influence. SELF-RAG scores a uniform 3/4, reflecting solid incremental work. Full review texts: [Mamba](examples/reviews/mamba_iclr2024_accepted.md) · [SELF-RAG](examples/reviews/selfrag_iclr2024_rejected.md).\n\nOutput folder:\n\n```text\n~/.clawphd/workspace/outputs/paper_review/<pdf_stem>/\n├── review.md # 6-section narrative + score table\n└── meta.json # venue, mode, scores, elapsed time\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "b3bb1ed934c200f89c0e804e6e232313903b20e4": {"t": 1785771355.690281, "n": 6345, "h": "af4e23810f5adcfe", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **AI Paper Review** — Peer-review any PDF against NeurIPS / ICLR / ICML / EuroSys / CVPR rubrics; produces narrative feedback + dimensional scores + Accept/Reject\n- [x] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter 3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n### AI Paper Review\n\nReview any paper PDF against real conference rubrics — produces a structured narrative review (Synopsis, Strengths, Weaknesses, Suggestions, References) plus venue-specific dimensional scores and an Accept/Reject recommendation:\n\n```bash\nexamples/ai_review_command.sh\n```\n\nSupported venues: NeurIPS · ICLR · ICML · EuroSys · OSDI · SOSP · CVPR · ICCV · General\n\n**Accuracy check — ICLR 2024 papers:**\n\n| Paper | Real outcome | Originality | Significance | Contribution | Overall | Decision |\n|-------|-------------|-------------|--------------|--------------|---------|----------|\n| [Mamba](https://arxiv.org/abs/2312.00752) (arXiv:2312.00752) | Accepted (Spotlight) | **4 / 4** | **4 / 4** | **4 / 4** | **8 / 10** | **Accept** |\n| [SELF-RAG](https://arxiv.org/abs/2310.11511) (arXiv:2310.11511) | Rejected from ICLR ¹ | 3 / 4 | 3 / 4 | 3 / 4 | 7 / 10 | Accept |\n\n> ¹ **Note on SELF-RAG:** This is a high-quality paper — it was accepted at EMNLP 2023 and has thousands of citations. Its ICLR 2024 rejection was due to a venue policy (already published elsewhere), **not** a reflection of paper quality. Our system correctly evaluates it as solid, well-executed research (7/10), just not at the same level of architectural breakthrough as Mamba (8/10, with three dimensions at 4/4).\n\nThe accepted paper (Mamba) scores higher on every breakthrough dimension. All three dimensions where Mamba achieves 4/4 — Originality, Significance, Contribution — are the strongest predictors of lasting architectural influence. SELF-RAG scores a uniform 3/4, reflecting solid incremental work. Full review texts: [Mamba](examples/reviews/mamba_iclr2024_accepted.md) · [SELF-RAG](examples/reviews/selfrag_iclr2024_rejected.md).\n\nOutput folder:\n\n```text\n~/.clawphd/workspace/outputs/paper_review/<pdf_stem>/\n├── review.md # 6-section narrative + score table\n└── meta.json # venue, mode, scores, elapsed time\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "536b0e074c817b5f2285345781c2def16d648075": {"t": 1785771356.923442, "n": 6538, "h": "fd8932300793488d", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **AI Paper Review** — Peer-review any PDF against NeurIPS / ICLR / ICML / EuroSys / CVPR rubrics; produces narrative feedback + dimensional scores + Accept/Reject\n- [x] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n### AI Paper Review\n\nReview any paper PDF against real conference rubrics — produces a structured narrative review (Synopsis, Strengths, Weaknesses, Suggestions, References) plus venue-specific dimensional scores and an Accept/Reject recommendation:\n\n```bash\nexamples/ai_review_command.sh\n```\n\nSupported venues: NeurIPS · ICLR · ICML · EuroSys · OSDI · SOSP · CVPR · ICCV · General\n\n**Accuracy check — ICLR 2024 papers:**\n\n| Paper | Real outcome | Originality | Significance | Contribution | Overall | Decision |\n|-------|-------------|-------------|--------------|--------------|---------|----------|\n| [Mamba](https://arxiv.org/abs/2312.00752) (arXiv:2312.00752) | Accepted (Spotlight) | **4 / 4** | **4 / 4** | **4 / 4** | **8 / 10** | **Accept** |\n| [SELF-RAG](https://arxiv.org/abs/2310.11511) (arXiv:2310.11511) | Rejected from ICLR ¹ | 3 / 4 | 3 / 4 | 3 / 4 | 7 / 10 | Accept |\n\n> ¹ **Note on SELF-RAG:** This is a high-quality paper — it was accepted at EMNLP 2023 and has thousands of citations. Its ICLR 2024 rejection was due to a venue policy (already published elsewhere), **not** a reflection of paper quality. Our system correctly evaluates it as solid, well-executed research (7/10), just not at the same level of architectural breakthrough as Mamba (8/10, with three dimensions at 4/4).\n\nThe accepted paper (Mamba) scores higher on every breakthrough dimension. All three dimensions where Mamba achieves 4/4 — Originality, Significance, Contribution — are the strongest predictors of lasting architectural influence. SELF-RAG scores a uniform 3/4, reflecting solid incremental work. Full review texts: [Mamba](examples/reviews/mamba_iclr2024_accepted.md) · [SELF-RAG](examples/reviews/selfrag_iclr2024_rejected.md).\n\nOutput folder:\n\n```text\n~/.clawphd/workspace/outputs/paper_review/<pdf_stem>/\n├── review.md # 6-section narrative + score table\n└── meta.json # venue, mode, scores, elapsed time\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "11b1bea98ec79febc3715c9a379ece39bc159007": {"t": 1785771358.2968163, "n": 6347, "h": "5ad18d0e78f008d9", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **AI Paper Review** — Peer-review any PDF against NeurIPS / ICLR / ICML / EuroSys / CVPR rubrics; produces narrative feedback + dimensional scores + Accept/Reject\n- [x] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n### AI Paper Review\n\nReview any paper PDF against real conference rubrics — produces a structured narrative review (Synopsis, Strengths, Weaknesses, Suggestions, References) plus venue-specific dimensional scores and an Accept/Reject recommendation:\n\n```bash\nexamples/ai_review_command.sh\n```\n\nSupported venues: NeurIPS · ICLR · ICML · EuroSys · OSDI · SOSP · CVPR · ICCV · General\n\n**Accuracy check — ICLR 2024 papers:**\n\n| Paper | Real outcome | Originality | Significance | Contribution | Overall | Decision |\n|-------|-------------|-------------|--------------|--------------|---------|----------|\n| [Mamba](https://arxiv.org/abs/2312.00752) (arXiv:2312.00752) | Accepted (Spotlight) | **4 / 4** | **4 / 4** | **4 / 4** | **8 / 10** | **Accept** |\n| [SELF-RAG](https://arxiv.org/abs/2310.11511) (arXiv:2310.11511) | Rejected from ICLR | 3 / 4 | 3 / 4 | 3 / 4 | 7 / 10 | Accept |\n\nThe accepted paper (Mamba) scores higher on every breakthrough dimension. All three dimensions where Mamba achieves 4/4 — Originality, Significance, Contribution — are the strongest predictors of lasting architectural influence. SELF-RAG scores a uniform 3/4, reflecting solid incremental work. Full review texts: [Mamba](examples/reviews/mamba_iclr2024_accepted.md) · [SELF-RAG](examples/reviews/selfrag_iclr2024_rejected.md).\n\n> SELF-RAG was rejected from ICLR 2024 because it had already been accepted at EMNLP 2023, not for quality reasons. Our system evaluates intrinsic paper quality and correctly ranks Mamba above SELF-RAG across all key dimensions.\n\nOutput folder:\n\n```text\n~/.clawphd/workspace/outputs/paper_review/<pdf_stem>/\n├── review.md # 6-section narrative + score table\n└── meta.json # venue, mode, scores, elapsed time\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "d0a9564d65112c4abb832ee3906e547a4bb2af8c": {"t": 1785771359.93301, "n": 4253, "h": "a41ecf82a8a19d35", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter 3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "d084b8e40f1b8ed0ca2866a4c824807ec2fc869f": {"t": 1785771361.3816931, "n": 4343, "h": "e2b9e5e39ef895c9", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter 3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "3d5a5de789e6a15b6e69ea5628fbca686ea88096": {"t": 1785771362.6444569, "n": 4446, "h": "9d8bf42a26700a7f", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [x] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "68dd12244003cc7611dc2d23cb35190a1204f040": {"t": 1785771364.0476253, "n": 4446, "h": "b0a9585e858e70f8", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "7ec82049edce1ceccdc769a967090c6552c0bda8": {"t": 1785771365.3606157, "n": 6380, "h": "e71532cf093d8e4e", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [x] **PDF → Markdown + Editable Figures** — Convert any paper PDF to structured Markdown; export all labelled figures as PNG + SVG + drawio\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n### PDF to Markdown + Editable Figures\n\nConvert a local paper PDF into structured Markdown and export figure assets (PNG + SVG + drawio, with editable rebuild fallback):\n\n```bash\nexamples/pdf2md_command.sh\n```\n\nTypical output folder:\n\n```text\n~/.clawphd/workspace/outputs/pdf2md/<pdf_stem>/\n```\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## Optional Dependencies\n\n### PDF → Markdown (`pdf_to_markdown` tool)\n\nCore pip packages (install with `pip install \"clawphd-ai[pdf2md]\"` or individually):\n\n```bash\npip install docling PyMuPDF Pillow\n```\n\n| Package | Purpose |\n|---|---|\n| `docling` | PDF → structured Markdown (default backend) |\n| `PyMuPDF` | Figure detection, PNG + SVG export via cropbox |\n| `Pillow` | PNG dimensions for drawio canvas sizing |\n\nOptional external CLI tools (provide higher-quality SVG or enable editable reconstruction):\n\n| Tool | Install | Purpose |\n|---|---|---|\n| `mutool` | [MuPDF](https://mupdf.com/releases/) → add to PATH | Best-quality vector SVG export per figure |\n| `pdf2svg` | `apt install pdf2svg` / [source](https://github.com/dawbarton/pdf2svg) | SVG export fallback |\n| `svgtodrawio` | `pip install svgtodrawio` or [source](https://github.com/mmunozba/svgtodrawio) | SVG → drawio conversion |\n\n**Editable reconstruction tools** (external research repos, not on PyPI):\n\nThese tools produce a fully editable `rebuild/rebuilt.svg` for each figure.\nWhen neither is installed, a built-in two-layer SVG fallback is used instead.\n\n**AutoFigure-Edit** (priority 1):\n```bash\ngit clone https://github.com/ResearAI/AutoFigure-Edit\ncd AutoFigure-Edit\npip install -e .\n# then tell ClawPhD where the entry script is:\nexport AUTOFIGURE_EDIT_CMD=\"python /path/to/AutoFigure-Edit/run.py\"\n```\n\n**Edit-Banana** (priority 2):\n```bash\ngit clone https://github.com/BIT-DataLab/Edit-Banana\ncd Edit-Banana\npip install -e .\nexport EDIT_BANANA_CMD=\"python /path/to/Edit-Banana/run.py\"\n```\n\nBoth env vars can also be set permanently in `~/.clawphd/config.json` or your shell profile.\n\n### Quick CLI test\n\n```bash\n# Smoke-test the pipeline against any local PDF (no rebuild tools required):\npython -m clawphd.agent.tools.pdf2md path/to/paper.pdf --no-rebuild\n\n# Full pipeline with rebuild fallback:\npython -m clawphd.agent.tools.pdf2md path/to/paper.pdf\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "a56b3107c0860eb754766106176a18d7d1b41d92": {"t": 1785771366.649072, "n": 4008, "h": "05a76edd356fccbd", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [x] **Figure Reference Extraction** — Search influential papers and extract real figures as editable SVG + PPTX\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n### Figure Reference Extraction\n\nThe Agent searches influential papers and extracts all labelled figures into an editable reference pack (PNG + SVG + PPTX):\n\n```bash\nexamples/figure_ref_command.sh\n```\n\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "6c5e19656e56ca3011ce5d6b4c1a3747967eebeb": {"t": 1785771367.9669793, "n": 3631, "h": "f3ef968c7607dd13", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n#### Iterative Refinement\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\nFirst Generation (Initial Output):\n\n\nAfter Multiple Self-feedback Iterations:\n\n1 iteration:\n\n3 iterations:\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "3d7243a6ca1dec35402e7f96c3a64577d8387fc4": {"t": 1785771370.471192, "n": 3625, "h": "2871f2d01f9a35fc", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n### Iterative Refinement with Human Feedback\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\n**First Generation (Initial Output):**\n\n\n**After Multiple Feedback Iterations:**\n\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "baaa70d0a9fc1baaba9a4ff4e3030a8b45c55f92": {"t": 1785771371.7717187, "n": 3589, "h": "2c5ddffa4bc78c6b", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [x] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n### Iterative Refinement with Human Feedback\n\nThe following images demonstrate the Agent's iterative refinement process for generating a HumanLLM framework diagram:\n\n**First Generation (Initial Output):**\n\n\n**After Multiple Feedback Iterations:**\n\n\n\nThese examples showcase how the Agent progressively improves diagram quality through human-in-the-loop feedback, resulting in more polished and publication-ready outputs.\n\n### Paper Website Generation\n\nThe Agent can turn academic papers into interactive web pages:\n\n```bash\nexamples/page_generation_command.sh\n```\n\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "81c5db41420f2036d52222c7d63e16e0c2eaabd7": {"t": 1785771373.108404, "n": 2849, "h": "8a3498880eac768a", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [ ] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\nGenerated results:\n\n\n\n\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "84ae651b285becdb23cd491ec95a1280916661e3": {"t": 1785771374.4774628, "n": 2831, "h": "04e0232cdc5e28c2", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [ ] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Examples\n### Diagram Generation\n```bash\nexamples/diagram_generation_command.sh\n```\n\n\n\n\n\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "79e73b1ecf594b85461de0ae7c232588b3438502": {"t": 1785771375.7274637, "n": 2651, "h": "0baadd9c478e7a5a", "c": "# 🐈 ClawPhD\n\nAn OpenClaw Agent for research that can turn academic papers into publication-ready diagrams, posters, videos, and more. This project is based on the nano version of OpenClaw: [Nanobot](https://github.com/HKUDS/nanobot).\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n[中文文档](README_CN.md)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [ ] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "cffb78ae85e9da470b80457ead212be861724531": {"t": 1785771377.0042067, "n": 2525, "h": "18b3fdb9b82062c0", "c": "# 🐈 ClawPhD\n\nA lightweight personal AI assistant framework with multi-channel support, tool use, and academic diagram generation.\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n## Features\n\n- [x] **Diagram Generation** — Create publication-quality academic illustrations and statistical plots from paper sections\n- [ ] **Paper Discovery** — Proactively search and summarize trending AI papers on a schedule\n- [ ] **Video Explainers** — Generate walkthrough videos from paper content\n- [ ] **Paper Websites** — Turn papers into interactive web pages\n- [ ] **Poster Generation** — Produce conference-ready posters from papers\n- [ ] **Code Synthesis** — Extract and generate reproducible code from paper methodologies\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}, "650743aac4d17b6d0785ba6a43689033f4207c52": {"t": 1785771378.1973963, "n": 2772, "h": "55bc35b8433686f7", "c": "# 🐈 ClawPhD\n\nA lightweight personal AI assistant framework with multi-channel support, tool use, and academic diagram generation.\n\n[](LICENSE)\n[](https://www.python.org/downloads/)\n\n## Features\n\n- **Multi-provider LLM** — Anthropic, OpenAI, Gemini, DeepSeek, Groq, OpenRouter, DashScope, Moonshot, vLLM, and more via [LiteLLM](https://github.com/BerriAI/litellm)\n- **Multi-channel chat** — Telegram, WhatsApp, Discord, Feishu/Lark\n- **Rich tool system** — File I/O, shell execution, web search/fetch, messaging, subagents, cron scheduling\n- **PaperBanana diagrams** — Generate publication-quality academic illustrations and statistical plots via Gemini or Replicate\n- **Skills** — Extensible markdown-based skill system (diagram-gen, GitHub, cron, summarize, weather, …)\n- **Memory** — Persistent long-term memory and daily notes\n- **Cron & Heartbeat** — Schedule recurring tasks; periodic self-checks\n- **Subagents** — Spawn background agents for long-running tasks\n\n## Quick Start\n\n### 1. Install\n\n```bash\n# From source\nuv pip install -e .\n\n# Or from PyPI\npip install clawphd-ai\n```\n\n### 2. Initialize\n\n```bash\nclawphd onboard\n```\n\nThis creates `~/.clawphd/config.json` and a default workspace at `~/.clawphd/workspace/`.\n\n### 3. Configure API Key\n\nEdit `~/.clawphd/config.json` and add at least one LLM provider key:\n\n```jsonc\n{\n \"providers\": {\n // Pick one (or more):\n \"openrouter\": { \"apiKey\": \"sk-or-...\" },\n \"anthropic\": { \"apiKey\": \"sk-ant-...\" },\n \"openai\": { \"apiKey\": \"sk-...\" },\n \"gemini\": { \"apiKey\": \"AI...\" },\n \"deepseek\": { \"apiKey\": \"sk-...\" }\n },\n \"agents\": {\n \"defaults\": {\n \"model\": \"anthropic/claude-opus-4-5\"\n }\n }\n}\n```\n\nFor PaperBanana diagram generation, also set the Replicate token:\n\n```bash\nexport REPLICATE_API_TOKEN=\"r8_...\"\n```\n\n### 4. Chat\n\n```bash\n# Single message\nclawphd agent -m \"Hello!\"\n\n# Interactive REPL\nclawphd agent\n```\n\n## CLI Reference\n\n| Command | Description |\n|---|---|\n| `clawphd onboard` | Initialize config and workspace |\n| `clawphd agent [-m MSG]` | Chat with the agent (interactive if no `-m`) |\n| `clawphd gateway [-p PORT]` | Start the multi-channel gateway |\n| `clawphd status` | Show config, API keys, and workspace status |\n| `clawphd channels status` | Show channel connection status |\n| `clawphd channels login` | Link WhatsApp via QR code |\n| `clawphd cron list` | List scheduled jobs |\n| `clawphd cron add` | Add a scheduled job (`--every`, `--cron`, or `--at`) |\n| `clawphd cron remove <ID>` | Remove a scheduled job |\n| `clawphd cron enable <ID>` | Enable / `--disable` a job |\n| `clawphd cron run <ID>` | Manually trigger a job |\n"}}}} |