Releases
platform-v1.2.60
Compare
Sorry, something went wrong.
No results found
1.2.60 (2026-06-08)
Features
add web crawler knowledge connector (#5326 ) (e02d65b )
chat: prompt-cache the stable prefix on Anthropic chat requests (#5346 ) (d713756 )
dev-stack: reuse persisted ports across restarts (#5404 ) (711430d )
environments: validation rule (allowlist regex) for MCP config values (#5406 ) (49b0485 )
llm-proxy: authenticate virtual API keys on Anthropic & OpenAI GET /models routes (#5400 ) (edd4ac6 )
llm: broaden best-model markers for the remaining providers (#5347 ) (9ed4267 )
skills: keep skill tools top-level and list skills in the prompt (#5403 ) (1cee2bc )
Bug Fixes
archestra-mcp: enforce agent tool assignment when executing tools (#5365 ) (2bc1d9b )
chat: reattach to the active run on duplicate-run 409 instead of dead-ending (#5363 ) (47f30e3 )
chat: stop spurious 'active response' blocks from orphaned chat runs (#5352 ) (b5ea6dd )
chat: tolerate trailing slash in provider model-fetcher base URLs (#5402 ) (34438cb )
mcp: steer agent recovery on tool/resource lookup misses (#5405 ) (a14e25f )
render mcp apps tool in tool search tools run_tool (#5366 ) (ffc67ba )
run-tool: recover from hallucinated tool names instead of a dead-end refusal (#5395 ) (eabb50e )
sandbox: close o11y, logging, and stale-permission gaps (#5359 ) (4bfffc6 )
skill-marketplace: install git in runtime image (#5367 ) (eb971a2 )
skills: clean-or-null skill description generation (#5371 ) (3457c5c )
skills: resolve service-account permissions in skill scope checks (#5332 ) (ba630f2 )
test: stabilize MSW integration tests (next-dev cold-compile timeout + blocked port) (#5349 ) (d9118c4 )
tilt: allow db-migrate to run in parallel (#5398 ) (eb4ed07 )
Performance Improvements
frontend: auto-select Turbopack for dev on Linux, webpack on macOS arm64 (#5377 ) (056ab2b )
Code Refactoring
run-tool: reuse assigned-tool set across pre-check and policy gate (#5397 ) (d3b48e6 )
Miscellaneous Chores
You can’t perform that action at this time.