{"repo": "calvaryyy/lance", "files": {"ava.py": {"43df7c63006a27aef4c5276e1cba288d0da8d7cf": {"t": 1785674063.0306737, "n": 15248, "h": "f296eeccc05b1da7", "c": "#!/usr/bin/env python3\n\"\"\"\nLANCE - AI Synthesis Engine\nPhase 3: Parse vuln_report.json, synthesize findings into a structured\nintelligence report with severity categorization and exploit correlation.\n\"\"\"\n\nimport json\nimport os\nimport sys\nimport re\nfrom pathlib import Path\nfrom datetime import datetime\n\n# ─────────────────────────────────────────────\n# CONFIG\n# ─────────────────────────────────────────────\n\nCONFIG_PATH = Path(__file__).parent / \"lance.conf\"\n\nSUPPORTED_PROVIDERS = [\"ollama\", \"openai\", \"gemini\", \"anthropic\", \"deepseek\"]\n\nSYSTEM_PROMPT = \"\"\"You are an expert penetration tester and vulnerability analyst.\nYou will be given a structured JSON report of vulnerabilities found during an\nexternal attack surface assessment. Your job is to synthesize this data into a\nclear, actionable intelligence report.\n\nYour report must:\n1. Summarize the overall security posture of the target\n2. List all findings grouped by severity (Critical → High → Medium → Low → Info)\n3. For each finding: explain what it is, why it matters, and the realistic exploit path\n4. Identify vulnerability chains — where multiple findings combine into a higher-impact attack\n5. Prioritize findings by a combined score of severity + ease of exploitability\n6. Provide concrete remediation steps per finding\n7. End with an executive summary of the top 3 most critical risks\n\nBe technical, precise, and direct. Do not pad the report. Do not repeat yourself.\nOutput in clean Markdown.\"\"\"\n\n# ─────────────────────────────────────────────\n# UTILITIES\n# ─────────────────────────────────────────────\n\ndef log(msg, level=\"INFO\"):\n    colors = {\n        \"INFO\":  \"\\033[94m\",\n        \"OK\":    \"\\033[92m\",\n        \"WARN\":  \"\\033[93m\",\n        \"ERROR\": \"\\033[91m\",\n        \"PHASE\": \"\\033[96m\",\n    }\n    reset = \"\\033[0m\"\n    prefix = colors.get(level, \"\") + f\"[{level}]\" + reset\n    print(f\"{prefix} {msg}\")\n\ndef load_config():\n    \"\"\"Load lance.conf into a dict. Returns empty dict if not found.\"\"\"\n    if not CONFIG_PATH.exists():\n        return {}\n    config = {}\n    with open(CONFIG_PATH) as f:\n        for line in f:\n            line = line.strip()\n            if not line or line.startswith(\"#\"):\n                continue\n            if \"=\" in line:\n                key, _, val = line.partition(\"=\")\n                config[key.strip()] = val.strip()\n    return config\n\ndef save_config(config):\n    \"\"\"Write config dict to lance.conf.\"\"\"\n    lines = [\"# LANCE Configuration File\\n# Auto-generated — edit with care\\n\"]\n    for key, val in config.items():\n        lines.append(f\"{key}={val}\")\n    with open(CONFIG_PATH, \"w\") as f:\n        f.write(\"\\n\".join(lines) + \"\\n\")\n    log(f\"Config saved → {CONFIG_PATH}\", \"OK\")\n\ndef load_vuln_report(report_path):\n    path = Path(report_path)\n    if not path.exists():\n        log(f\"Vuln report not found: {report_path}\", \"ERROR\")\n        sys.exit(1)\n    with open(path) as f:\n        return json.load(f)\n\ndef prompt_ai_provider():\n    \"\"\"Interactive prompt if user hasn't specified --ai flag.\"\"\"\n    print(\"\\n\\033[96m[LANCE] No AI provider specified.\\033[0m\")\n    print(\"Select a provider:\\n\")\n    print(\"  1) Ollama     (local inference — no API key needed)\")\n    print(\"  2) OpenAI     (GPT-4o, GPT-4-turbo, etc.)\")\n    print(\"  3) Gemini     (Google Gemini Pro)\")\n    print(\"  4) Anthropic  (Claude)\")\n    print(\"  5) DeepSeek   (DeepSeek-V3, R1)\")\n    print()\n    while True:\n        choice = input(\"Enter choice [1-5]: \").strip()\n        mapping = {\"1\": \"ollama\", \"2\": \"openai\", \"3\": \"gemini\", \"4\": \"anthropic\", \"5\": \"deepseek\"}\n        if choice in mapping:\n            return mapping[choice]\n        print(\"Invalid choice. Enter a number between 1 and 5.\")\n\ndef prompt_ollama_model():\n    \"\"\"Let user pick an installed Ollama model.\"\"\"\n    try:\n        import subprocess\n        result = subprocess.run(\"ollama list\", shell=True, capture_output=True, text=True)\n        lines = result.stdout.strip().splitlines()\n        models = []\n        for line in lines[1:]:  # skip header\n            parts = line.split()\n            if parts:\n                models.append(parts[0])\n    except Exception:\n        models = []\n\n    if not models:\n        log(\"No Ollama models found. Run: ollama pull <model>\", \"WARN\")\n        model = input(\"Enter model name manually (e.g. llama3, mistral): \").strip()\n        return model or \"llama3\"\n\n    print(\"\\nInstalled Ollama models:\")\n    for i, m in enumerate(models, 1):\n        print(f\"  {i}) {m}\")\n    print()\n    while True:\n        choice = input(f\"Select model [1-{len(models)}]: \").strip()\n        try:\n            idx = int(choice) - 1\n            if 0 <= idx < len(models):\n                return models[idx]\n        except ValueError:\n            pass\n        print(\"Invalid choice.\")\n\n\n# ─────────────────────────────────────────────\n# PROVIDER CLIENTS\n# ─────────────────────────────────────────────\n\ndef call_ollama(prompt, model):\n    try:\n        import requests\n    except ImportError:\n        log(\"requests not installed. Run: pip install requests --break-system-packages\", \"ERROR\")\n        sys.exit(1)\n\n    log(f\"Calling Ollama [{model}]...\", \"INFO\")\n    try:\n        response = requests.post(\n            \"http://localhost:11434/api/generate\",\n            json={\"model\": model, \"prompt\": prompt, \"stream\": False},\n            timeout=300\n        )\n        response.raise_for_status()\n        return response.json().get(\"response\", \"\")\n    except Exception as e:\n        log(f\"Ollama error: {e}\", \"ERROR\")\n        sys.exit(1)\n\n\ndef call_openai(prompt, api_key, model=\"gpt-4o\"):\n    try:\n        import requests\n    except ImportError:\n        log(\"requests not installed. Run: pip install requests --break-system-packages\", \"ERROR\")\n        sys.exit(1)\n\n    log(f\"Calling OpenAI [{model}]...\", \"INFO\")\n    headers = {\"Authorization\": f\"Bearer {api_key}\", \"Content-Type\": \"application/json\"}\n    payload = {\n        \"model\": model,\n        \"messages\": [\n            {\"role\": \"system\", \"content\": SYSTEM_PROMPT},\n            {\"role\": \"user\", \"content\": prompt}\n        ],\n        \"max_tokens\": 4096,\n    }\n    try:\n        response = requests.post(\n            \"https://api.openai.com/v1/chat/completions\",\n            headers=headers, json=payload, timeout=120\n        )\n        response.raise_for_status()\n        return response.json()[\"choices\"][0][\"message\"][\"content\"]\n    except Exception as e:\n        log(f\"OpenAI error: {e}\", \"ERROR\")\n        sys.exit(1)\n\n\ndef call_gemini(prompt, api_key, model=\"gemini-1.5-pro\"):\n    try:\n        import requests\n    except ImportError:\n        log(\"requests not installed.\", \"ERROR\")\n        sys.exit(1)\n\n    log(f\"Calling Gemini [{model}]...\", \"INFO\")\n    url = f\"https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={api_key}\"\n    payload = {\n        \"contents\": [{\"parts\": [{\"text\": SYSTEM_PROMPT + \"\\n\\n\" + prompt}]}],\n        \"generationConfig\": {\"maxOutputTokens\": 4096}\n    }\n    try:\n        response = requests.post(url, json=payload, timeout=120)\n        response.raise_for_status()\n        return response.json()[\"candidates\"][0][\"content\"][\"parts\"][0][\"text\"]\n    except Exception as e:\n        log(f\"Gemini error: {e}\", \"ERROR\")\n        sys.exit(1)\n\n\ndef call_anthropic(prompt, api_key, model=\"claude-sonnet-4-20250514\"):\n    try:\n        import requests\n    except ImportError:\n        log(\"requests not installed.\", \"ERROR\")\n        sys.exit(1)\n\n    log(f\"Calling Anthropic [{model}]...\", \"INFO\")\n    headers = {\n        \"x-api-key\": api_key,\n        \"anthropic-version\": \"2023-06-01\",\n        \"Content-Type\": \"application/json\"\n    }\n    payload = {\n        \"model\": model,\n        \"max_tokens\": 4096,\n        \"system\": SYSTEM_PROMPT,\n        \"messages\": [{\"role\": \"user\", \"content\": prompt}]\n    }\n    try:\n        response = requests.post(\n            \"https://api.anthropic.com/v1/messages\",\n            headers=headers, json=payload, timeout=120\n        )\n        response.raise_for_status()\n        return response.json()[\"content\"][0][\"text\"]\n    except Exception as e:\n        log(f\"Anthropic error: {e}\", \"ERROR\")\n        sys.exit(1)\n\n\ndef call_deepseek(prompt, api_key, model=\"deepseek-chat\"):\n    try:\n        import requests\n    except ImportError:\n        log(\"requests not installed.\", \"ERROR\")\n        sys.exit(1)\n\n    log(f\"Calling DeepSeek [{model}]...\", \"INFO\")\n    headers = {\"Authorization\": f\"Bearer {api_key}\", \"Content-Type\": \"application/json\"}\n    payload = {\n        \"model\": model,\n        \"messages\": [\n            {\"role\": \"system\", \"content\": SYSTEM_PROMPT},\n            {\"role\": \"user\", \"content\": prompt}\n        ],\n        \"max_tokens\": 4096,\n    }\n    try:\n        response = requests.post(\n            \"https://api.deepseek.com/v1/chat/completions\",\n            headers=headers, json=payload, timeout=120\n        )\n        response.raise_for_status()\n        return response.json()[\"choices\"][0][\"message\"][\"content\"]\n    except Exception as e:\n        log(f\"DeepSeek error: {e}\", \"ERROR\")\n        sys.exit(1)\n\n\n# ─────────────────────────────────────────────\n# PROMPT BUILDER\n# ─────────────────────────────────────────────\n\ndef build_synthesis_prompt(vuln_report):\n    \"\"\"Build the prompt fed to the AI from the vuln report.\"\"\"\n    meta = vuln_report.get(\"meta\", {})\n    summary = vuln_report.get(\"summary\", {})\n    findings = vuln_report.get(\"findings\", [])\n\n    # Slim down findings to essential fields to stay within context limits\n    slim_findings = []\n    for f in findings:\n        slim_findings.append({\n            \"name\": f.get(\"name\"),\n            \"severity\": f.get(\"severity\"),\n            \"url\": f.get(\"url\"),\n            \"category\": f.get(\"category\"),\n            \"description\": f.get(\"description\", \"\")[:300],\n            \"tool\": f.get(\"tool\"),\n            \"references\": f.get(\"references\", [])[:3],\n            \"extra\": {k: v for k, v in f.get(\"extra\", {}).items()\n                      if k in (\"template_id\", \"plugin\", \"fixed_in\", \"matcher_name\")},\n        })\n\n    prompt = f\"\"\"You are analyzing the following vulnerability assessment report.\n\nTARGET: {meta.get('target', 'Unknown')}\nSCAN DATE: {meta.get('timestamp', 'Unknown')}\nASSETS SCANNED: {summary.get('assets_scanned', 0)}\nTOTAL FINDINGS: {summary.get('total_findings', 0)}\n\nSEVERITY BREAKDOWN:\n{json.dumps(summary.get('severity_breakdown', {}), indent=2)}\n\nFINDINGS:\n{json.dumps(slim_findings, indent=2)}\n\n---\nGenerate a full penetration test intelligence report following your instructions.\nFocus especially on vulnerability chains and prioritized remediation.\n\"\"\"\n    return prompt\n\n\n# ─────────────────────────────────────────────\n# REPORT WRITER\n# ─────────────────────────────────────────────\n\ndef save_synthesis_report(content, output_dir, target):\n    safe_target = target.replace(\"/\", \"_\").replace(\":\", \"_\").replace(\".\", \"_\")\n    timestamp = datetime.utcnow().strftime(\"%Y%m%d_%H%M%S\")\n    md_path = output_dir / f\"synthesis_{safe_target}_{timestamp}.md\"\n    with open(md_path, \"w\") as f:\n        f.write(f\"# LANCE Intelligence Report\\n\")\n        f.write(f\"**Target:** {target}  \\n\")\n        f.write(f\"**Generated:** {datetime.utcnow().isoformat()}Z  \\n\")\n        f.write(f\"**Engine:** LANCE AI Synthesis v0.1\\n\\n\")\n        f.write(\"---\\n\\n\")\n        f.write(content)\n    log(f\"Synthesis report saved → {md_path}\", \"OK\")\n    return md_path\n\n\n# ─────────────────────────────────────────────\n# ENTRY POINT\n# ─────────────────────────────────────────────\n\ndef run_synthesis(vuln_report_path, provider=None, model_override=None):\n    log(f\"Loading vuln report: {vuln_report_path}\", \"INFO\")\n    vuln_report = load_vuln_report(vuln_report_path)\n    config = load_config()\n    output_dir = Path(vuln_report_path).parent\n\n    # Resolve provider\n    if not provider:\n        provider = config.get(\"default_provider\")\n    if not provider:\n        provider = prompt_ai_provider()\n\n    provider = provider.lower()\n    if provider not in SUPPORTED_PROVIDERS:\n        log(f\"Unknown provider: {provider}. Choose from: {', '.join(SUPPORTED_PROVIDERS)}\", \"ERROR\")\n        sys.exit(1)\n\n    log(f\"Provider: {provider}\", \"INFO\")\n\n    # Build prompt\n    prompt = build_synthesis_prompt(vuln_report)\n\n    # Dispatch to provider\n    if provider == \"ollama\":\n        model = model_override or config.get(\"ollama_model\")\n        if not model:\n            model = prompt_ollama_model()\n            config[\"ollama_model\"] = model\n            save_config(config)\n        result = call_ollama(prompt, model)\n\n    else:\n        # API key providers\n        key_field = f\"{provider}_api_key\"\n        api_key = config.get(key_field)\n\n        if not api_key:\n            log(f\"No API key found for {provider}.\", \"WARN\")\n            log(f\"Run: python3 lance_config.py --set {key_field} <your_key>\", \"INFO\")\n            api_key = input(f\"Enter {provider.upper()} API key (will be saved to config): \").strip()\n            if not api_key:\n                log(\"No API key provided. Exiting.\", \"ERROR\")\n                sys.exit(1)\n            config[key_field] = api_key\n            save_config(config)\n\n        default_models = {\n            \"openai\":    \"gpt-4o\",\n            \"gemini\":    \"gemini-1.5-pro\",\n            \"anthropic\": \"claude-sonnet-4-20250514\",\n            \"deepseek\":  \"deepseek-chat\",\n        }\n        model = model_override or config.get(f\"{provider}_model\") or default_models[provider]\n\n        dispatch = {\n            \"openai\":    lambda: call_openai(prompt, api_key, model),\n            \"gemini\":    lambda: call_gemini(prompt, api_key, model),\n            \"anthropic\": lambda: call_anthropic(prompt, api_key, model),\n            \"deepseek\":  lambda: call_deepseek(prompt, api_key, model),\n        }\n        result = dispatch[provider]()\n\n    if not result:\n        log(\"AI returned empty response.\", \"ERROR\")\n        sys.exit(1)\n\n    log(\"Synthesis complete.\", \"OK\")\n    report_path = save_synthesis_report(result, output_dir, vuln_report[\"meta\"].get(\"target\", \"unknown\"))\n\n    print(\"\\n\" + \"─\" * 50)\n    print(f\"  SYNTHESIS COMPLETE\")\n    print(\"─\" * 50)\n    print(f\"  Provider : {provider} [{model}]\")\n    print(f\"  Report   : {report_path}\")\n    print(\"─\" * 50 + \"\\n\")\n\n    return report_path\n\n\nif __name__ == \"__main__\":\n    import argparse\n\n    parser = argparse.ArgumentParser(\n        description=\"LANCE AI Synthesis — Intelligent vulnerability report generation\",\n        formatter_class=argparse.RawDescriptionHelpFormatter,\n        epilog=\"\"\"\nExamples:\n  python3 ai_synthesis.py -r results/example.com/vuln_report.json --ai ollama\n  python3 ai_synthesis.py -r results/example.com/vuln_report.json --ai gemini\n  python3 ai_synthesis.py -r results/example.com/vuln_report.json --ai anthropic --model claude-opus-4-20250514\n        \"\"\"\n    )\n    parser.add_argument(\"-r\", \"--report\", required=True, help=\"Path to vuln_report.json from Phase 2\")\n    parser.add_argument(\"--ai\", dest=\"provider\", choices=SUPPORTED_PROVIDERS,\n                        help=\"AI provider to use for synthesis\")\n    parser.add_argument(\"--model\", dest=\"model_override\", default=None,\n                        help=\"Override the default model for the selected provider\")\n\n    args = parser.parse_args()\n    run_synthesis(args.report, provider=args.provider, model_override=args.model_override)\n"}}}}