main
docs: readme
docs: add chat_model_orig parameter for Azure OpenAI API v1 (#943)
fix(batch): stop_on_first_result should wait for the first non-None result (#1072)
docs: fix bullet-pt syntax
docs: correct what full_eval=True actually disables
chore: rename crawl-4-ai extra to crawl4ai
docs: azure-openai-models.md
Docagent hypoq (#683)
docs: update file-input.md
Add Vertex AI support for Gemini models via configurable api_base (#960)
docs: glhf.chat
docs: handle llm no tool
tests: add crawl4ai extra in pytest CI
docs: notes, FAQ on knowledge-graphs
docs: further notes on langdb
feat: _retain_raw_result flag in ToolMessage; implements #605 (#606)
feat: litellm-proxy/ prefix to switch in/out of litellm-proxy -notest (#797)
feat: add robust llama.cpp embedding response format handling (#922)
fix: carry LLM "no content" faithfully as None (fixes Gemini 3.x tool calls) (#1062)
docs: update llm-pdf-parser.md to render correctly [skip-ci] (#857)
docs: marker-pdf.md update on docling conflict
feat: MarkitdownDocxParser (#795)
feat(mcp): map JSON-Schema enum/const→Literal and anyOf/oneOf→Union for tool params (#1059)
Add recognize_recipient_in_content config option for recipient routing (#961)
feat: implement client caching to prevent resource exhaustion (#883)
Support async client factory for OpenAI HTTP client configuration (#990)
feat: LLMConfig.async_stream_quiet flag (#575)
refactor: improve DocChatAgent retrieval configuration and fix RRF bias (#874)
Add Portkey AI Gateway support to Langroid (#859)
feat: Complete Pydantic V2 Migration (#901)
fix: QdrantDB lock file issue - add close() and context manager support (#879)
docs: quiet mode
Add reasoning parameter to LLM response callbacks (#965)
Support short-lived / rotating credentials via api_key_provider (Vertex AI, Azure AD) (#1081)
Add Seltz web search integration with tool and examples (#994)
docs: struc output typo
feat(TaskTool): Add agent_name parameter for custom sub-agent naming
fix: avoid empty msg in ChatDocument.to_LLMMessage (#822)
update docs
Add Xquik MCP search example (#1024)
docs: url_loader, chunking
docs: typos