v0.2.1 fix(mcp): live reader on the MCP pipeline + config model routing + Vertex preflight#173
Merged
Merged
Conversation
… routing + Vertex preflight (0.2.1) Three first-run fixes from real plugin feedback (the MCP/plugin surface, not just the CLI): - The MCP research pipeline now resolves its reader via config.resolve_corpus_reader() (the live Arctic Shift API default) instead of a hardcoded HF ArcticReader at four mcp/tools.py sites. 0.1.1's live-reader default swapped the CLI + facade but missed the MCP path, so plugin runs pulled posts from HuggingFace and 429'd while preflight reported active_reader=arctic_shift_api. Now they agree. - A config `model` that is a routable ref (e.g. "deepseek/deepseek-v4-flash") routes on its own without also pinning `provider`, before key-order/Vertex autodetection — so setting `model` alone works (a pinned provider+model is unchanged; a bare id still can't self-route). - preflight/doctor now flag (error) GOOGLE_GENAI_USE_VERTEXAI on + google extra missing, the exact landmine that silently routes embeddings to a missing Vertex SDK. +6 regression tests. Clean-room full suite 1139 passed (the local failures are stray ~/.config/metalworks + ~/clique/metalworks.toml deepseek configs and a populated store). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Workflows to automatically generate PRs for you. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
0.2.1 — first-run fixes for the MCP/plugin surface
From real plugin feedback: a user on a Vertex-configured machine hit (1) HuggingFace 429s on the post pull and (2) the planner ignoring their
modelconfig. Root causes, all on the MCP surface that the CLI/SDK fixes in 0.1.1 missed.Fixed
mcp/tools.pyhardcodedArcticReader(probe_sleep_s=0.0)(HF Parquet) at four sites (_build_deps,research_plan_brief,arctic_list_months,arctic_pull_threads) — 0.1.1's live-reader default swapped the CLI + facade but never the MCP path. So MCP/plugin runs pulled posts from HF and 429'd, whilepreflight(which calls the resolver) reportedactive_reader: arctic_shift_api. All four now callconfig.resolve_corpus_reader(); preflight and the pipeline finally agree.modelthat's a routable ref works on its own.model = "deepseek/deepseek-v4-flash"without a pinnedproviderwas silently ignored for routing (fell through to key-order/Vertex), so a machine with strayGOOGLE_GENAI_USE_VERTEXAIresolved Google. It now routes by vendor namespace before autodetection. Pinnedprovider+modelunchanged; a bare id still can't self-route.Added
preflight/doctorflag the Vertex landmine (error):GOOGLE_GENAI_USE_VERTEXAIon butgoogle.genainot installed → chat→Vertex and embeddings both fail on the missing SDK even with an OpenRouter key. The exact first-run failure the preamble exists to catch.Verification
tests/test_firstrun_021.py): MCP_build_depswires the resolver; config-model-alone routing (and the bare-id and provider+model cases); the Vertex preflight hint.~/.config/metalworks+~/clique/metalworks.tomldeepseek configs and a populated store — absent in CI).On merge: tag
v0.2.1→ PyPI.🤖 Generated with Claude Code