Add qx-labs/agents-deep-research

agents/qx-labs__agents-deep-research/README.md

@@ -0,0 +1,69 @@
+# agents-deep-research
+
+> A multi-agent deep research assistant powered by the [OpenAI Agents SDK](https://github.com/openai/openai-agents-python).
+
+## What it does
+
+**agents-deep-research** performs in-depth, iterative research on any topic and produces a comprehensive, well-cited Markdown report — autonomously, without asking clarifying questions.
+
+It ships two research modes:
+
+| Mode | Best for | Typical output |
+|---|---|---|
+| `IterativeResearcher` | Focused single-topic queries | Up to ~5 pages / 1 000 words |
+| `DeepResearcher` | Structured, multi-section reports | 20+ pages with parallel section research |
+
+## How it works
+
+The agent pipeline runs in a loop:
+
+1. **Knowledge Gap Agent** — analyses current research state and surfaces the next unknown.
+2. **Tool Selector Agent** — decides which specialised agents to dispatch (web search, site crawler) and crafts precise 3-6 word queries.
+3. **Tool Agents** — execute in parallel, returning structured findings with citations.
+4. **Thinking / Observations Agent** — reflects on findings, notes contradictions, and steers the next loop iteration.
+5. **Writer Agent** — synthesises everything into a final Markdown report with numbered URL citations.
+
+For the `DeepResearcher`, a **Planner Agent** first creates a report outline, then parallel `IterativeResearcher` instances tackle each section, and a **Proofreader Agent** unifies the whole document.
+
+## Key capabilities
+
+- **Model agnostic** — works with OpenAI, Anthropic, Gemini, DeepSeek, Perplexity, OpenRouter, Azure OpenAI, Hugging Face, Ollama, LM Studio — any OpenAI-API-compatible model with structured output support.
+- **Extensible tools** — custom tool agents can be added by dropping a file into `deep_researcher/agents/tool_agents/`.
+- **Autonomous** — no mid-run clarifying questions; runs fully unattended.
+- **Traced** — optional OpenAI trace monitoring per research session.
+- **Installable** — available as `pip install deep-researcher`.
+
+## Example usage
+
+```bash
+# Long-form structured report
+deep-researcher --mode deep \
+  --query "Comprehensive analysis of the EV battery supply chain" \
+  --max-iterations 5 --max-time 20 --verbose
+
+# Quick focused report
+deep-researcher --mode simple \
+  --query "Latest EU AI Act compliance requirements" \
+  --max-iterations 3 --max-time 10 --output-length "3 pages"
+```
+
+```python
+import asyncio
+from deep_researcher import DeepResearcher, IterativeResearcher
+
+# Structured multi-section report
+researcher = DeepResearcher(max_iterations=5, max_time_minutes=20)
+report = asyncio.run(researcher.run("Impact of quantum computing on cryptography"))
+
+# Quick focused report
+researcher = IterativeResearcher(max_iterations=3, max_time_minutes=5)
+report = asyncio.run(researcher.run("UK renewable energy policy 2024-2025", output_length="2 pages"))
+print(report)
+```
+
+## Links
+
+- **Repository:** https://github.com/qx-labs/agents-deep-research
+- **PyPI:** https://pypi.org/project/deep-researcher/
+- **Author:** Jai Juneja @ [QX Labs](https://www.qxlabs.com)
+- **License:** Apache-2.0

agents/qx-labs__agents-deep-research/metadata.json

@@ -0,0 +1,15 @@
+{
+  "name": "agents-deep-research",
+  "author": "qx-labs",
+  "description": "Multi-agent iterative deep research assistant. Identifies knowledge gaps, selects tools, runs parallel web searches, and synthesises comprehensive cited reports.",
+  "repository": "https://github.com/qx-labs/agents-deep-research",
+  "path": "",
+  "version": "1.0.0",
+  "category": "research",
+  "tags": ["deep-research", "multi-agent", "web-search", "report-generation", "openai-agents-sdk", "iterative", "autonomous", "llm", "python"],
+  "license": "Apache-2.0",
+  "model": "gpt-4o",
+  "adapters": ["openai", "system-prompt"],
+  "icon": false,
+  "banner": false
+}