1 line
7.0 KiB
JSON
1 line
7.0 KiB
JSON
{"repo": "chetangoel01/cognitive-assistant", "files": {"app/config.py": {"5e04059b6903348207a8897146bac055c4d29a90": {"t": 1785747495.5205247, "n": 2264, "h": "8c550899caa6f9cc", "c": "\"\"\"Central configuration, loaded from environment / .env.\"\"\"\nfrom __future__ import annotations\n\nfrom functools import lru_cache\n\nfrom pydantic_settings import BaseSettings, SettingsConfigDict\n\n\nclass Settings(BaseSettings):\n model_config = SettingsConfigDict(\n env_file=\".env\", env_file_encoding=\"utf-8\", extra=\"ignore\"\n )\n\n # LLM (OpenRouter)\n openrouter_api_key: str = \"\"\n openrouter_base_url: str = \"https://openrouter.ai/api/v1\"\n llm_model: str = \"deepseek/deepseek-chat\"\n llm_fast_model: str = \"deepseek/deepseek-chat\"\n llm_temperature: float = 0.2\n\n # Database\n database_url: str = \"postgresql://postgres@localhost:5432/cognitive_assistant\"\n use_pgvector: bool = False\n\n # Local embeddings (model2vec static embeddings - pure NumPy, no native runtime).\n # For multilingual retrieval, swap to \"minishlab/M2V_multilingual_output\".\n embedding_model: str = \"minishlab/potion-base-8M\"\n embedding_dim: int = 256\n # Reranking is done by the LLM (single batched call). No local cross-encoder\n # needed, which keeps the stack free of onnxruntime/torch.\n enable_rerank: bool = True\n\n # Retrieval / chunking\n chunk_size_tokens: int = 256\n chunk_overlap_tokens: int = 48\n vector_top_k: int = 20\n keyword_top_k: int = 20\n rerank_top_n: int = 6\n\n # Agentic behaviour\n enable_query_decomposition: bool = True\n enable_self_check: bool = True\n enable_judge: bool = True\n memory_turns: int = 3\n\n # Reasoning (\"thinking\") tokens, surfaced live in the UI agent trace.\n # Requested via OpenRouter's unified `reasoning` parameter; models without\n # reasoning support (e.g. deepseek-chat) silently ignore it, while models\n # like deepseek-r1 stream their chain of thought.\n enable_thinking: bool = True\n thinking_effort: str = \"medium\" # low | medium | high\n\n # Server\n host: str = \"127.0.0.1\"\n port: int = 8000\n\n @property\n def llm_configured(self) -> bool:\n key = self.openrouter_api_key\n return bool(\n key\n and key.startswith(\"sk-\")\n and \"...\" not in key # reject the .env.example placeholder\n and len(key) >= 24\n )\n\n\n@lru_cache\ndef get_settings() -> Settings:\n return Settings()\n"}, "3e9d0ec300c52c1ea3e880e9dc265d791022b270": {"t": 1785773587.1234233, "n": 2264, "h": "8c550899caa6f9cc", "c": "\"\"\"Central configuration, loaded from environment / .env.\"\"\"\nfrom __future__ import annotations\n\nfrom functools import lru_cache\n\nfrom pydantic_settings import BaseSettings, SettingsConfigDict\n\n\nclass Settings(BaseSettings):\n model_config = SettingsConfigDict(\n env_file=\".env\", env_file_encoding=\"utf-8\", extra=\"ignore\"\n )\n\n # LLM (OpenRouter)\n openrouter_api_key: str = \"\"\n openrouter_base_url: str = \"https://openrouter.ai/api/v1\"\n llm_model: str = \"deepseek/deepseek-chat\"\n llm_fast_model: str = \"deepseek/deepseek-chat\"\n llm_temperature: float = 0.2\n\n # Database\n database_url: str = \"postgresql://postgres@localhost:5432/cognitive_assistant\"\n use_pgvector: bool = False\n\n # Local embeddings (model2vec static embeddings - pure NumPy, no native runtime).\n # For multilingual retrieval, swap to \"minishlab/M2V_multilingual_output\".\n embedding_model: str = \"minishlab/potion-base-8M\"\n embedding_dim: int = 256\n # Reranking is done by the LLM (single batched call). No local cross-encoder\n # needed, which keeps the stack free of onnxruntime/torch.\n enable_rerank: bool = True\n\n # Retrieval / chunking\n chunk_size_tokens: int = 256\n chunk_overlap_tokens: int = 48\n vector_top_k: int = 20\n keyword_top_k: int = 20\n rerank_top_n: int = 6\n\n # Agentic behaviour\n enable_query_decomposition: bool = True\n enable_self_check: bool = True\n enable_judge: bool = True\n memory_turns: int = 3\n\n # Reasoning (\"thinking\") tokens, surfaced live in the UI agent trace.\n # Requested via OpenRouter's unified `reasoning` parameter; models without\n # reasoning support (e.g. deepseek-chat) silently ignore it, while models\n # like deepseek-r1 stream their chain of thought.\n enable_thinking: bool = True\n thinking_effort: str = \"medium\" # low | medium | high\n\n # Server\n host: str = \"127.0.0.1\"\n port: int = 8000\n\n @property\n def llm_configured(self) -> bool:\n key = self.openrouter_api_key\n return bool(\n key\n and key.startswith(\"sk-\")\n and \"...\" not in key # reject the .env.example placeholder\n and len(key) >= 24\n )\n\n\n@lru_cache\ndef get_settings() -> Settings:\n return Settings()\n"}, "76af55a5ca4cb2e14a2c1c027e1594af29c5898c": {"t": 1785773588.9175146, "n": 1886, "h": "157018134d4c55da", "c": "\"\"\"Central configuration, loaded from environment / .env.\"\"\"\nfrom __future__ import annotations\n\nfrom functools import lru_cache\n\nfrom pydantic_settings import BaseSettings, SettingsConfigDict\n\n\nclass Settings(BaseSettings):\n model_config = SettingsConfigDict(\n env_file=\".env\", env_file_encoding=\"utf-8\", extra=\"ignore\"\n )\n\n # LLM (OpenRouter)\n openrouter_api_key: str = \"\"\n openrouter_base_url: str = \"https://openrouter.ai/api/v1\"\n llm_model: str = \"deepseek/deepseek-chat\"\n llm_fast_model: str = \"deepseek/deepseek-chat\"\n llm_temperature: float = 0.2\n\n # Database\n database_url: str = \"postgresql://postgres@localhost:5432/cognitive_assistant\"\n use_pgvector: bool = False\n\n # Local embeddings (model2vec static embeddings - pure NumPy, no native runtime).\n # For multilingual retrieval, swap to \"minishlab/M2V_multilingual_output\".\n embedding_model: str = \"minishlab/potion-base-8M\"\n embedding_dim: int = 256\n # Reranking is done by the LLM (single batched call). No local cross-encoder\n # needed, which keeps the stack free of onnxruntime/torch.\n enable_rerank: bool = True\n\n # Retrieval / chunking\n chunk_size_tokens: int = 256\n chunk_overlap_tokens: int = 48\n vector_top_k: int = 20\n keyword_top_k: int = 20\n rerank_top_n: int = 6\n\n # Agentic behaviour\n enable_query_decomposition: bool = True\n enable_self_check: bool = True\n enable_judge: bool = True\n memory_turns: int = 3\n\n # Server\n host: str = \"127.0.0.1\"\n port: int = 8000\n\n @property\n def llm_configured(self) -> bool:\n key = self.openrouter_api_key\n return bool(\n key\n and key.startswith(\"sk-\")\n and \"...\" not in key # reject the .env.example placeholder\n and len(key) >= 24\n )\n\n\n@lru_cache\ndef get_settings() -> Settings:\n return Settings()\n"}}}} |