v3.6.0

[assafelovic]

Introducing a super stable version with enhanced improvements across security, stability and performance. We're also introducing a collaboration with Langchain with the latest deep agents framework here - https://github.com/assafelovic/gpt-researcher/tree/main/deep_agents

Thank you and welcome to all the new contributors!

What's Changed

• security: sanitize untrusted content, add SECURITY.md, pin brotli by @assafelovic in security: sanitize untrusted content, add SECURITY.md, pin brotli #1820
• Fix silently dropped token limits, real usage-based cost tracking, and research pipeline robustness by @assafelovic in Fix silently dropped token limits, real usage-based cost tracking, and research pipeline robustness #1861
• Add Deep Agents example: GPT Researcher as the research engine in a LangChain deep agent by @assafelovic in Add Deep Agents example: GPT Researcher as the research engine in a LangChain deep agent #1857
• Resolve post-merge conflicts: 4 retriever/context fixes (supersedes fix(context): map url to source on compressor fast path #1895-1898) by @assafelovic in Resolve post-merge conflicts: 4 retriever/context fixes (supersedes #1895-1898) #1906
• Retriever hardening: 20 guard fixes against malformed results (supersedes fix(retriever): return empty list instead of None from SerperSearch.search #1837…fix(retriever): skip redundant Tavily MCP when direct Tavily is active #1890) by @assafelovic in Retriever hardening: 20 guard fixes against malformed results (supersedes #1837…#1890) #1907
• Scraper robustness: 6 fixes (title/PDF-detection/temp-file/dimensions) (supersedes fix(scraper): return empty string from extract_title for empty title tags #1824…fix(scraper): remove PyMuPDF temp file even when PDF loading fails #1842) by @assafelovic in Scraper robustness: 6 fixes (title/PDF-detection/temp-file/dimensions) (supersedes #1824…#1842) #1908
• Multi-agent robustness: bound revision loops + exact sentinels (supersedes fix(multi_agents): exact None sentinel for reviewer and fact-checker #1883/fix(multi_agents): bound reviewer/reviser draft revision loop #1885/fix(multi_agents): bound fact-checker writer revision loop #1886) by @assafelovic in Multi-agent robustness: bound revision loops + exact sentinels (supersedes #1883/#1885/#1886) #1909
• Core/misc hardening: 11 fixes (costs/query/llm/agent/mcp/config) (supersedes fix(costs): fall back to default encoding for non-OpenAI embedding models #1822…fix(config): coerce none/null/empty for Optional env fields #1902) by @assafelovic in Core/misc hardening: 11 fixes (costs/query/llm/agent/mcp/config) (supersedes #1822…#1902) #1910
• Reconcile master→main: GetXAPI retriever, Claude 4.x temp fix, MiniMax M3, curate_sources fixes by @assafelovic in Reconcile master→main: GetXAPI retriever, Claude 4.x temp fix, MiniMax M3, curate_sources fixes #1911
• feat: add Nebius Token Factory as LLM and embedding provider by @teionarr in feat: add Nebius Token Factory as LLM and embedding provider #1891
• chore(multi_agents): nest AG2 variant under multi_agents/ag2 by @assafelovic in chore(multi_agents): nest AG2 variant under multi_agents/ag2 #1912
• docs(skill): fix drift in .claude reference docs (retrievers, multi-agents, api) by @assafelovic in docs(skill): fix drift in .claude reference docs (retrievers, multi-agents, api) #1913

New Contributors

• @teionarr made their first contribution in feat: add Nebius Token Factory as LLM and embedding provider #1891

Full Changelog: v3.5.1...v3.6.0

---

This discussion was created from the release v3.6.0.

[teionarr]

awesome 🎉 thanks for the shoutout, @assafelovic. happy to keep contributing