v0.4.0: Phase 2 surface expansion (framework adapters + learnings + analytics + gen-platform)#8
Merged
Merged
Conversation
… Tier B generators)
Phase 2 of the expansion roadmap. v0.4.0 adds the framework adapter
layer, a learning store + outcome aggregator, an LLM-call telemetry
aggregator with cost estimation, and config generators for the
remaining Tier B native platforms.
New code surfaces, each force-included into the wheel:
_adapters/lib -> vstack.adapters (LangChain / LangGraph /
CrewAI / AutoGen /
LlamaIndex / Pydantic AI /
OpenAI / Anthropic /
Open WebUI tool bindings)
_learnings/lib -> vstack.learnings (LearningStore +
vstack-learn CLI)
_analytics/lib -> vstack.analytics (FileTelemetrySink +
TelemetryAggregator +
CostEstimator +
vstack-analytics CLI)
Plus _memory/lib/_platforms.py adds 14 platform generators behind
`vstack-config gen-platform`: cursor, cline, continue, roo-code,
windsurf, zed, aider, goose, kiro, openclaw, codex-cli, opencode,
docker-compose, claude-desktop. --write + --out + --force land the
snippet at the suggested path.
Framework adapters share a single registry-driven dispatcher
(run_pattern_dispatch). Same input model + same detection output as
the MCP server and REST API; the difference is just the framework
wrapper. Adding a pattern to the registry adds it to every adapter.
Packaging:
- 5 new optional extras: [langchain], [langgraph], [crewai],
[llamaindex], [pydantic_ai]. [adapters] bundles them; [all]
bundles everything including the framework adapters.
- 2 new [project.scripts]: vstack-learn, vstack-analytics.
- Force-include extended; pytest testpaths extended.
CI:
- mypy strict loop covers _adapters / _learnings / _analytics.
- Test job installs langchain-core / langgraph / llama-index-core /
pydantic-ai so framework-gated tests actually run in CI.
- Lint job covers new dirs.
- Release smoke test imports vstack.adapters / vstack.learnings /
vstack.analytics.
Tests:
- +84 new tests (51 adapters + 15 learnings + 15 analytics + 7
gen-platform). Suite total 1,895 passing (up from 1,811 in
v0.3.0). Mypy strict clean across all 7 surface lib dirs.
Version bump 0.3.0 -> 0.4.0; runtime __version__ paired with the
pyproject bump.
The Phase 2 mypy + security jobs install langchain-core / langgraph /
llama-index-core / pydantic-ai so framework-gated adapter tests run.
That unmasks two unrelated upstream churn issues:
1. mypy strict on the pre-existing AAR OpenAIClient flags
'messages=[{...}]' as incompatible with the newer openai SDK's
stricter TypedDict-based type stubs (which the new install
pulls). Add type: ignore[arg-type] on the two affected lines;
the runtime behaviour is unchanged.
2. pip-audit picks up transitive CVEs in pydantic-ai 0.4.3,
python-dotenv 1.1.1, uv 0.9.30, diskcache 5.6.3 -- all pulled
from framework dep trees we don't control. The workflow's
fallback pass without --strict was meant to be warn-only but
pip-audit still exits 1 when vulns exist. Append '|| echo ...'
so the step's exit code becomes the echo's success. bandit
continues to gate security on vstack's own code.
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
1 participant
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.
Summary
Phase 2 of the expansion roadmap. Lands the framework adapter layer + learning store + telemetry aggregator + Tier B platform generators.
vstack.adapters— registry-driven wrappers for LangChain / LangGraph / CrewAI / AutoGen / LlamaIndex / Pydantic AI / OpenAI Assistants / Anthropic Messages / Open WebUI. Same 34 patterns, framework-native shapes.vstack.learnings+vstack-learn— append-only JSONL store at~/.vstack/learnings.jsonlwith outcome aggregation.vstack.analytics+vstack-analytics—FileTelemetrySinkfor the existingrecord_llm_callevents plus per-pattern / per-model / per-day rollups + cost estimation.vstack-config gen-platform— 14 ready-to-paste config snippets for the non-MCP-default clients (cursor / cline / continue / roo-code / windsurf / zed / aider / goose / kiro / openclaw / codex-cli / opencode / docker-compose / claude-desktop).Bumps 0.3.0 → 0.4.0. New extras:
[langchain],[langgraph],[crewai],[llamaindex],[pydantic_ai],[adapters].Test plan
pytest -q— 1,895 passing (was 1,811 in v0.3.0; +84 new)ruff check .+ruff format --check .cleanmypy --strict --ignore-missing-importsclean on all 7 surface lib dirs (_mcp,_memory,_upgrade,_api,_adapters,_learnings,_analytics)vstack-config gen-platform cursor/--list/--write --out/ refusing to overwrite without--force