Phases 142-148: HTTP/MCP parity sweep (workflow, analysis flags, narrate/explain, guide, watch, graph, insights)#124
Merged
Conversation
CLI's `gitsema watch` has add/run/list/remove; HTTP previously only exposed POST /watch/add and POST /watch/run. Add GET /watch (list, newest first) and DELETE /watch/:name (remove by name, 404 if not found), mirroring the CLI's watch list/watch remove exactly — name is used as the identifier since that's the saved_queries table's unique key and what the CLI's own `watch remove <name>` deletes by. Updates docs/parity.md (new watch row in the Tool Matrix, which had none before), docs/features.md, README.md, docs/PLAN.md (Phase 146 marked complete with the :name-vs-:id deviation noted), adds integration tests to tests/serverRoutes.test.ts, and adds a changeset. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
POST /api/v1/guide/chat now accepts lens: 'semantic'|'structural'|'hybrid', mirroring CLI `guide --lens` byte-for-byte by appending the same tool-bias hint suffix to the question before calling runGuide(). The remote multi-turn/session half of Phase 145's original scope (HTTP equivalent of CLI guide --interactive) is a genuine open design question rather than a flag gap, and is explicitly deferred to a follow-up phase rather than bolted on speculatively — documented in docs/PLAN.md's Phase 145 entry and written up as a full design-gaps entry in docs/feature-ideas.md. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
POST /analysis/workflow's WorkflowBodySchema previously only accepted 3 of the 8 productized workflow templates the CLI's `workflow run` supports. Extends the template enum to all 8 (pr-review, incident, release-audit, onboarding, ownership-intel, arch-drift, knowledge-portal, regression-forecast), wiring each new template to the same core functions the CLI calls (all already imported/used elsewhere in analysis.ts, so no follow-up gaps). Adds role/ref body fields generally (mirroring CLI --role/--ref); base was already present and un-gated, and is intentionally left as a no-op matching the CLI's own dead --base flag. Updates docs/PLAN.md (Phase 142 marked complete with deviations), docs/parity.md, docs/features.md, and README.md; adds a changeset. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
…HTTP routes Phase 144: POST /narrate and POST /explain gain an `evidenceOnly` body field mirroring the CLI's safe-by-default `--narrate`/`--evidence-only` toggle (HTTP callers previously had no way to request it explicitly), plus a structured `evidence` array in both responses. POST /explain additionally gains `log` (error/stack-trace context) and `files` (search-scope glob, mirrored as-is including the CLI's pre-existing files-doesn't-filter-commits gap), and both routes gain `lens`, which drives a `structuralContext` response field on /explain when combined with `files` under a structural/hybrid lens (no-op on /narrate, which has no per-file enrichment target). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
Close a grab-bag of small, independent HTTP-vs-CLI flag gaps across src/server/routes/analysis.ts found by the Interface Parity Track audit: - merge-audit: add `base` (merge-base override, mirrors CLI --base) - merge-preview: add top/iterations/edgeThreshold/enhancedKeywordsN, plus useEnhancedLabels (deviation — enhancedKeywordsN is a no-op without it) - branch-summary: add enhancedLabels/enhancedKeywordsN, which now slice nearestConcepts[].topKeywords in the JSON response (this route's only "view"), since computeBranchSummary() has no enhanced-labels concept of its own — the CLI only uses these flags for text-mode display - clusters: add iterations/edgeThreshold/enhancedKeywordsN - security-scan: add highConfidenceOnly (filters to confidence === 'high') - impact: add chunks/level/lens — lens=structural|hybrid makes this route a thin blast-radius alias, closing a real divergence where HTTP previously only ever did semantic-lens impact analysis - semantic-diff: add hybrid/bm25Weight, which required fixing a pre-existing CLI bug where `diff`'s --hybrid/--bm25-weight flags were declared but never wired to anything; computeSemanticDiff() gained an optional candidateBlobs parameter (mirrors computeAuthorContributions's Phase-141 pattern) so both interfaces now genuinely blend BM25 via hybridSearch() - semantic-blame: add level (file/symbol) as an alternate spelling of the pre-existing searchSymbols boolean Updates docs/PLAN.md (Phase 143 marked complete with deviations noted), docs/parity.md (§1/§2.2/§6), and docs/features.md. Adds unit tests for computeSemanticDiff's new candidateBlobs param and HTTP integration tests for every new flag across all 8 routes. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
Adds POST /api/v1/graph/{callers,callees,neighbors,path,relate,similar,
unused,cycles,deps,co-change,blast-radius} HTTP routes and MCP tools
graph_path/graph_relate/graph_similar/graph_unused/cycles/deps/co_change/
blast_radius, closing the pre-existing HTTP/MCP gap for most of the CLI's
graph command family (previously only `hotspots` had a route, and only
call_graph/graph_neighbors/hotspots had MCP tools). All are thin
schema-validated wrappers over the already-tested src/core/graph/* query
helpers. graph build stays CLI-only: it truncates and rebuilds
graph_nodes/edges (mutating index-maintenance, not a query), matching the
existing precedent for index vacuum/gc/rebuild-fts/etc.
Updates docs/parity.md (matrix + roadmap + flag notes), docs/features.md,
README.md, CLAUDE.md's MCP tool inventory, and docs/PLAN.md's Phase 147
entry with the graph-build design rationale. Adds HTTP route tests to
tests/serverRoutes.test.ts and a minor changeset.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
…e rest Adds MCP tools (src/mcp/tools/insights.ts) and POST /api/v1/insights/* HTTP routes (src/server/routes/insights.ts) for the 10 CLI commands that genuinely had zero HTTP/MCP exposure: bisect, refactor-candidates, lifecycle, cherry-pick-suggest, file-diff, pr-report, regression-gate, code-review, map, and heatmap. regression-gate and code-review had their core logic extracted out of the CLI layer into src/core/search/regressionGate.ts and src/core/search/codeReview.ts so CLI/MCP/HTTP share one implementation. cross-repo-similarity, project, and ci-diff were judged CLI-shaped by nature (raw local filesystem paths, local precompute step, and CI-runner/GitHub-comment side effects respectively) and intentionally left CLI-only; diff was found to be the exact same implementation as the already-exposed semantic-diff, not a separate feature. Full triage rationale is in docs/parity.md's Phase 148 footnotes. Also fixes a pre-existing bug in computeRefactorCandidates' default level: 'symbol' query, which referenced a nonexistent symbol_embeddings.blob_hash column and threw on any index with symbol embeddings present. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WyC4ii9yHPz7bsXcJLbedN
…-deploy-142-148-dhe3mq Phase 142 — workflow HTTP route parity
…-deploy-142-148-dhe3mq Phase 143 — Analysis-route small-fixes bundle # Conflicts: # docs/PLAN.md # docs/parity.md
…-deploy-142-148-dhe3mq Phase 144 — narrate/explain HTTP routes evidence-only toggle and missing context flags # Conflicts: # docs/parity.md
…-deploy-142-148-dhe3mq Phase 145 — guide HTTP route lens support (multi-turn session deferred) # Conflicts: # docs/parity.md
…-deploy-142-148-dhe3mq Phase 146 — watch list/remove HTTP routes # Conflicts: # tests/serverRoutes.test.ts
…-deploy-142-148-dhe3mq Phase 147 — Graph command family HTTP/MCP exposure # Conflicts: # docs/features.md # docs/parity.md
…-deploy-142-148-dhe3mq Phase 148 — Triage remaining zero-HTTP/MCP-exposure CLI commands # Conflicts: # CLAUDE.md # docs/features.md # docs/parity.md
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
resolveRefToTimestamp() shells out to `git log <ref>` when a ref isn't a parseable date, but CI checkouts are shallow/detached-HEAD with no local `main` branch — only HEAD resolves. Switch the two new bisect tests to a date string for goodRef (resolveRefToTimestamp tries Date-parsing first). Also closes session.rawDb before rmSync() in evolutionBranch.test.ts's afterEach, per CLAUDE.md's Windows-CI testing convention (unrelated pre-existing gap, but blocking this PR's Windows CI job).
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.
Summary
Implements PLAN.md Phases 142-148, closing the remaining CLI/HTTP/MCP interface-parity gaps found in the audit that started with Phase 138:
POST /analysis/workflownow supports all 8 CLI workflow templates (was 3), plusrole/refbody fields.merge-auditbase,merge-preview/clusterscluster-tuning flags,security-scanhighConfidenceOnly,impactchunks/level/lens,semantic-diffhybrid/bm25Weight(fixing a pre-existing dead-flag bug in the CLI's owndiffcommand along the way),semantic-blamelevel.narrate/explainHTTP routes gain anevidenceOnlytoggle (safe-by-default parity with the CLI),explaingainslog/files, both gainlens.guideHTTP route gainslens; remote multi-turn/session support is deliberately deferred (open design question, documented indocs/feature-ideas.md).GET /watch/DELETE /watch/:nameHTTP routes (previously onlyadd/runexisted).callers,callees,neighbors,path,relate,similar,unused,cycles,deps,co-change,blast-radius);graph buildstays CLI-only by design (mutating index-maintenance op).bisect,refactor-candidates,lifecycle,cherry-pick-suggest,file-diff,pr-report,regression-gate,code-review,map,heatmap); 3 correctly excluded as CLI-shaped or redundant (cross-repo-similarity,project,ci-diff/diff).MCP tool count is now 56 (up from 38 at the start of this track). Each phase was built independently in an isolated worktree and merged sequentially; conflicts (mostly in
docs/PLAN.md,docs/parity.md,docs/features.md,CLAUDE.md, andtests/serverRoutes.test.ts) were reconciled by hand, including correcting cumulative MCP tool-count arithmetic across phases.Docs updated:
docs/PLAN.md(all 7 phases marked complete with deviations noted),docs/features.md,README.md,docs/parity.md,CLAUDE.md. 7 changesets added (allminor).Test plan
pnpm build— cleanpnpm test— full suite green (1572 passed, 22 skipped, 0 failed)tests/serverRoutes.test.ts,tests/insightsRoutes.test.ts,tests/insightsTools.test.ts,tests/serverGuideRoute.test.ts)Generated by Claude Code