fix: route count_tokens to api.anthropic.com, not proxy base_url

feat(memory): resolve Qdrant connection from HEADROOM_QDRANT_* env vars (#31)

Merge pull request #262 from gglucass/fix/traffic-learner-error-recovery

Merge pull request #246 from Kayzo/fix/memory-batch-onnx-sqlitevec

test(memory): raise patch coverage from 83% to 98% on error_recovery fixes

fix(memory): collapse and decay error_recovery patterns in MEMORY.md

fix(memory): batch onnx embeddings and sqlite-vec ops

test(learn): cover flush_to_file, backend edge cases, and hydrate/bump error paths

fix(learn): persist real evidence_count and bump on re-sighting

feat(learn): live flush of traffic patterns to agent-native context files

chore(memory): add EXTERNAL backend extension points

Fix ruff lint errors in test files

feat: add live traffic learning + cross-agent memory writers (--learn flag)

feat(router): adaptive compression with Read lifecycle and context-pressure scaling

Fix CI: guard starlette imports, asyncio.run(), deprecate datetime.utcnow()

Add centralized ML model configuration

Fix test to expect VectorBackend.AUTO as default

Add multi-provider memory system with auto-detection

fix(tests): skip memory tests when hnswlib not available

Add DiffCompressor and fix hnswlib SIGILL crash on CI

Fix asyncio event loop error in Python 3.10+ tests

Add global httpx.ReadTimeout handler for memory tests

Fix mypy errors and add network timeout handler for flaky CI tests

Add hierarchical memory system with graph + vector storage

Make hnswlib optional and skip tests when unavailable

Replace legacy memory system with HierarchicalMemory

Add persistent memory system with zero-latency inline extraction