feat(repo)!: retire the summarizer package and LLM symbol-summary feature

CLAUDE.md

@@ -11,7 +11,7 @@ tiers.
 - `impact` — dependents of a target up to a configurable depth, with a risk tier.
 - `detect_changes` — map an uncommitted or committed diff to affected symbols.
 - `list_findings` — browse SARIF findings from the latest scan by severity and rule.
-- `sql` — read-only SQL against the local temporal store (cochanges + symbol_summaries), 5 s timeout; the node/edge graph is queried via the typed tools or Cypher via the MCP `sql` tool.
+- `sql` — read-only SQL against the local temporal store (cochanges), 5 s timeout; the node/edge graph is queried via the typed tools or Cypher via the MCP `sql` tool.
 
 Run `codehub analyze` after pulling new commits so the index stays aligned
 with the working tree. `codehub status` reports staleness.
@@ -84,7 +84,7 @@ provides a `code-analyst` subagent and 11 skills. Install via
 
 The entire index lives in ONE `<repo>/.codehub/store.sqlite` file (WAL),
 via Node's built-in `node:sqlite` — graph nodes, edges, embeddings, and
-the temporal tables (cochanges, symbol summaries, the
+the temporal tables (cochanges and the
 `codehub query --sql` escape hatch). One `SqliteStore` class implements
 **both** `IGraphStore` and `ITemporalStore`; `openStore()` returns that
 single instance as both the `graph` and `temporal` views, so call sites

commitlint.config.mjs

@@ -54,7 +54,6 @@ export default {
         "scip-ingest",
         "search",
         "storage",
-        "summarizer",
         "wiki",
         "plugin",
         "deps",

packages/cli/src/agent-context.ts

@@ -38,7 +38,7 @@ tiers.
 - \`impact\` — dependents of a target up to a configurable depth, with a risk tier.
 - \`detect_changes\` — map an uncommitted or committed diff to affected symbols.
 - \`list_findings\` — browse SARIF findings from the latest scan by severity and rule.
-- \`sql\` — read-only SQL against the local temporal store (cochanges + symbol_summaries), 5 s timeout; the node/edge graph is queried via the typed tools or Cypher via the MCP \`sql\` tool.
+- \`sql\` — read-only SQL against the local temporal store (cochanges), 5 s timeout; the node/edge graph is queried via the typed tools or Cypher via the MCP \`sql\` tool.
 
 Run \`codehub analyze\` after pulling new commits so the index stays aligned
 with the working tree. \`codehub status\` reports staleness.

packages/cli/src/commands/analyze.test.ts

@@ -1,15 +1,5 @@
 /**
  * Unit tests for `codehub analyze` helpers.
- *
- * P04 deliverables covered here:
- *   - `resolveMaxSummariesCap` arithmetic for `--max-summaries auto`:
- *     prior-run seed → floor(count × 0.1) clamped at 500; first-run
- *     fallback → 50; explicit numbers pass through; summaries disabled
- *     → 0 regardless of input.
- *   - `CODEHUB_BEDROCK_DISABLED=1` env kill-switch semantics: when the
- *     caller short-circuits `summariesEnabled` to false (mirroring how
- *     `runAnalyze` and the CLI entry point honor the env var) the cap
- *     collapses to 0 even under `auto`.
  */
 
 import assert from "node:assert/strict";
@@ -25,10 +15,8 @@ import {
   detectCoverageReport,
   isWorkingTreeDirty,
   resolveCoverageEnabled,
-  resolveMaxSummariesCap,
   resolveSbomEnabled,
   resolveScanEnabled,
-  resolveSummariesEnabled,
 } from "./analyze.js";
 import { selectScannerIds } from "./scan.js";
 import { computeScanFingerprint, shouldSkipScan } from "./scan-fingerprint.js";
@@ -88,135 +76,6 @@ async function initGitRepo(dir: string): Promise<string> {
   return headPromise;
 }
 
-test("resolveMaxSummariesCap: auto resolves to floor(count × 0.1) when seed is known", async () => {
-  // 1234 callables → 10% = 123.4 → floor = 123.
-  const cap = await resolveMaxSummariesCap("/unused", "auto", true, async () => 1234);
-  assert.equal(cap, 123);
-});
-
-test("resolveMaxSummariesCap: auto clamps at the 500 cap for large repos", async () => {
-  // 10_000 callables would be 1000 uncapped; the policy clamps at 500.
-  const cap = await resolveMaxSummariesCap("/unused", "auto", true, async () => 10_000);
-  assert.equal(cap, 500);
-});
-
-test("resolveMaxSummariesCap: auto falls back to 50 on first run (no prior seed)", async () => {
-  // `undefined` models "no prior store at the expected path".
-  const cap = await resolveMaxSummariesCap("/unused", "auto", true, async () => undefined);
-  assert.equal(cap, 50);
-});
-
-test("resolveMaxSummariesCap: undefined defaults to the auto path", async () => {
-  // A caller that doesn't pass --max-summaries at all gets the same
-  // behavior as explicit `auto`.
-  const cap = await resolveMaxSummariesCap("/unused", undefined, true, async () => 200);
-  assert.equal(cap, 20);
-});
-
-test("resolveMaxSummariesCap: explicit numbers pass through unchanged", async () => {
-  const cap = await resolveMaxSummariesCap("/unused", 42, true, async () => 999);
-  assert.equal(cap, 42);
-});
-
-test("resolveMaxSummariesCap: negative integers clamp to 0 (dry-run)", async () => {
-  const cap = await resolveMaxSummariesCap("/unused", -5, true, async () => 999);
-  assert.equal(cap, 0);
-});
-
-test("resolveMaxSummariesCap: non-integer numbers are floored", async () => {
-  const cap = await resolveMaxSummariesCap(
-    "/unused",
-    7.9 as unknown as number,
-    true,
-    async () => 0,
-  );
-  assert.equal(cap, 7);
-});
-
-test("resolveMaxSummariesCap: summariesEnabled=false collapses to 0 regardless of input", async () => {
-  // This is the behavior chain the CLI depends on when either
-  // `--no-summaries` or `CODEHUB_BEDROCK_DISABLED=1` short-circuits
-  // `summariesEnabled` upstream.
-  const autoCap = await resolveMaxSummariesCap("/unused", "auto", false, async () => 1_000);
-  assert.equal(autoCap, 0, "auto + disabled must dry-run");
-  const numericCap = await resolveMaxSummariesCap("/unused", 100, false, async () => 1_000);
-  assert.equal(numericCap, 0, "explicit cap + disabled must dry-run");
-});
-
-test("resolveMaxSummariesCap: seed of 0 yields a cap of 0 (no callables to summarize)", async () => {
-  const cap = await resolveMaxSummariesCap("/unused", "auto", true, async () => 0);
-  assert.equal(cap, 0);
-});
-
-test("resolveMaxSummariesCap: seed of 5 yields a cap of 0 under the 10% rule", async () => {
-  // 5 × 0.1 = 0.5 → floor = 0. Tiny repos dry-run until they grow.
-  const cap = await resolveMaxSummariesCap("/unused", "auto", true, async () => 5);
-  assert.equal(cap, 0);
-});
-
-// ---------------------------------------------------------------------------
-// resolveSummariesEnabled — fast-default contract: LLM summaries are opt-in.
-// `codehub analyze` runs tree-sitter + SCIP + cochange phases only by default,
-// so a fresh invocation never spends on Bedrock or blocks on a network hop.
-// ---------------------------------------------------------------------------
-
-test("resolveSummariesEnabled: default-off when both env and flag are absent", () => {
-  assert.equal(resolveSummariesEnabled(undefined, {}), false);
-});
-
-test("resolveSummariesEnabled: explicit --summaries opts in", () => {
-  assert.equal(resolveSummariesEnabled(true, {}), true);
-});
-
-test("resolveSummariesEnabled: explicit --no-summaries stays off", () => {
-  assert.equal(resolveSummariesEnabled(false, {}), false);
-});
-
-test("resolveSummariesEnabled: CODEHUB_BEDROCK_SUMMARIES=1 opts in (env-only)", () => {
-  // Operators can enable summaries for a whole CI job without editing every
-  // invocation. Only the literal "1" triggers — anything else is treated as
-  // absent, mirroring the kill-switch semantics below.
-  assert.equal(resolveSummariesEnabled(undefined, { CODEHUB_BEDROCK_SUMMARIES: "1" }), true);
-  assert.equal(resolveSummariesEnabled(undefined, { CODEHUB_BEDROCK_SUMMARIES: "0" }), false);
-  assert.equal(resolveSummariesEnabled(undefined, { CODEHUB_BEDROCK_SUMMARIES: "" }), false);
-});
-
-test("resolveSummariesEnabled: CODEHUB_BEDROCK_DISABLED=1 kills the phase", () => {
-  assert.equal(resolveSummariesEnabled(undefined, { CODEHUB_BEDROCK_DISABLED: "1" }), false);
-});
-
-test("resolveSummariesEnabled: kill-switch wins over --summaries=true", () => {
-  // Operator passed --summaries explicitly but the env var forces off.
-  // Required so CI / restricted environments can lock out Bedrock without
-  // auditing every invocation site.
-  assert.equal(resolveSummariesEnabled(true, { CODEHUB_BEDROCK_DISABLED: "1" }), false);
-});
-
-test("resolveSummariesEnabled: kill-switch wins over CODEHUB_BEDROCK_SUMMARIES=1", () => {
-  // Both env vars set → disable wins. This lets a CI environment pin the
-  // opt-in globally while still allowing per-job kill-switch overrides.
-  assert.equal(
-    resolveSummariesEnabled(undefined, {
-      CODEHUB_BEDROCK_SUMMARIES: "1",
-      CODEHUB_BEDROCK_DISABLED: "1",
-    }),
-    false,
-  );
-});
-
-test("resolveSummariesEnabled: --no-summaries wins over CODEHUB_BEDROCK_SUMMARIES=1", () => {
-  // Explicit CLI false beats env opt-in. Matches how --no-flag usually
-  // wins against ambient config everywhere else in the CLI.
-  assert.equal(resolveSummariesEnabled(false, { CODEHUB_BEDROCK_SUMMARIES: "1" }), false);
-});
-
-test("resolveSummariesEnabled: CODEHUB_BEDROCK_DISABLED=0 does not enable the phase", () => {
-  // Only the literal "1" on the opt-in var flips this; anything else leaves
-  // summaries in their (fast, off) default.
-  assert.equal(resolveSummariesEnabled(undefined, { CODEHUB_BEDROCK_DISABLED: "0" }), false);
-  assert.equal(resolveSummariesEnabled(undefined, { CODEHUB_BEDROCK_DISABLED: "" }), false);
-});
-
 // ---------------------------------------------------------------------------
 // Dirty-tree bypass on the analyze fast-path.
 // ---------------------------------------------------------------------------