fix: fast 3-agent web demo (Surveyor->Planner->Formatter, ~60s) (#12)

fix: limit web chat to 2 iterations for interactive response time (#11)

fix: make stream_chat async to avoid event loop conflict with Gradio (#10)

docs: rewrite README to match actual architecture and codebase (#9)

fix: add missing Critic→Orchestrator edge for completeness critiques (#8)

feat(ci): add test suite and run tests in CI before deploy (#7)

fix: render D3 architecture in iframe (Gradio strips scripts) (#6)

feat: add animated D3.js architecture diagram to Архітектура tab (#5)

fix: use eu-central-1 Bedrock model ID (eu.anthropic.claude-sonnet-4-6) (#4)

fix: remove gr.load() proxy, always build local app (#3)

fix(ci): update HF Space URL after rename to lmaf (#2)

refactor: rename LegalIntern to LMAF (Legal Multi-Agent Framework) (#1)

feat: switch to AWS Bedrock only, remove direct Anthropic/OpenAI API

fix: remove type param from ChatInterface (Gradio 6.x compat)

fix: remove unsupported type param from gr.Chatbot (Gradio 6.x)

feat: chat interface with real multi-agent pipeline

fix(ci): split deploy into separate steps, fix exit code handling

fix(ci): use docker exec for health check (no host port mapping)

fix(ci): pull SecondLayer repo on prod before docker build

fix(ci): use self-hosted runner + SSH to prod for deploy

fix(ci): remove invalid script_stop parameter from ssh-action

ci: add deploy workflow (prod SSH + HF Space sync)

fix: add src to sys.path for HF Space imports

add requirements.txt for HF Space

feat: initial LegalIntern multi-agent consultation system