chore(recall): remove the proactive recall hook

README.md

@@ -339,10 +339,7 @@ This plugin captures session activity and stores it in your Deeplake workspace:
 | `HIVEMIND_SUMMARY_EVERY_HOURS` | `2`                  | Time-based summary cadence, used when at least one new event has arrived since the last summary. |
 | `HIVEMIND_WIKI_WORKER`    | _(none)_                  | Set to `1` to disable the background session-summary worker entirely (no `claude -p` summary runs). Also set automatically inside the worker as a recursion guard. Capture and recall keep working. |
 | `HIVEMIND_GRAPH_ON_STOP`  | _(none)_                  | Set to `0` to disable the code-graph rebuild that runs on `Stop` / `SessionEnd`. |
-| `HIVEMIND_EMBEDDINGS`     | `true`                    | Set to `false` to force lexical-only mode  |
-| `HIVEMIND_PROACTIVE_RECALL_DISABLED` | _(none)_       | Set to `1` to disable **proactive recall** (auto-searching team memory on each recall-worthy prompt and injecting a relevant snippet into the agent's context). On by default. Does **not** affect capture or the agent's own grep/skill recall. Alt form: `HIVEMIND_PROACTIVE_RECALL=0`. |
-| `HIVEMIND_RECALL_MIN_OVERLAP` | `2`                   | Proactive recall (lexical mode): min distinct prompt keywords a summary must share to be injected. Higher = stricter. |
-| `HIVEMIND_RECALL_TIMEOUT_MS` | `1000`                 | Proactive recall: hard cap on the synchronous search path; on timeout it skips rather than delay the turn. |
+| `HIVEMIND_EMBEDDINGS`     | _(none)_                  | Read once, when `~/.deeplake/config.json` has no `embeddings.enabled` yet: unset or `false` seeds it off, any other value (`true`, `1`, ...) seeds it on. Afterwards only `hivemind embeddings install`/`enable` (persist on) and `disable`/`uninstall` (persist off) change it. |
 | `HIVEMIND_DEBUG`          | _(none)_                  | Set to `1` for verbose hook debug logs     |
 
 ## Per-directory config (`.hivemind`)
@@ -439,18 +436,6 @@ Hivemind ships with a local embedding daemon (nomic-embed-text-v1.5) for hybrid
 
 Full guide: **[docs/EMBEDDINGS.md](docs/EMBEDDINGS.md)**.
 
-## Proactive recall
-
-On a recall-worthy prompt (errors, "how did we…", substantive requests — acks and short follow-ups are skipped), Hivemind automatically searches the team's summaries and, if the top hit clears a relevance bar, injects one attributed snippet (`recalled from <teammate> · <date>`) into the agent's context — so prior work shows up *unprompted*, not only when the agent decides to search. Semantic when embeddings are installed, otherwise lexical (ILIKE keyword overlap), so it works without the embedding model. The search is latency-bounded and skips silently on any miss or error.
-
-**On by default.** To turn it off (capture and the agent's own grep/skill recall are unaffected):
-
-```bash
-HIVEMIND_PROACTIVE_RECALL_DISABLED=1 claude   # or HIVEMIND_PROACTIVE_RECALL=0
-```
-
-Tune precision/latency with `HIVEMIND_RECALL_MIN_OVERLAP` and `HIVEMIND_RECALL_TIMEOUT_MS` (see the table above). Every recall-worthy invocation is recorded to `~/.deeplake/recall-events.jsonl` for usage/hit-rate analysis.
-
 ## Summaries
 
 After each session, a background worker generates an AI-written wiki summary and stores it in the `memory` table alongside its 768-dim embedding. Long sessions checkpoint mid-session every 50 messages or 2 hours (configurable). The wiki worker shells out to the host agent's own CLI (`claude -p`, `codex exec`, `pi --print`, …) so no separate API key is needed. Browse summaries at `~/.deeplake/memory/summaries/`.

esbuild.config.mjs

@@ -80,7 +80,6 @@ const ccHooks = [
   { entry: "dist/src/hooks/session-start-setup.js", out: "session-start-setup" },
   { entry: "dist/src/hooks/session-notifications.js", out: "session-notifications" },
   { entry: "dist/src/hooks/capture.js", out: "capture" },
-  { entry: "dist/src/hooks/recall.js", out: "recall" },
   { entry: "dist/src/hooks/pre-tool-use.js", out: "pre-tool-use" },
   { entry: "dist/src/hooks/session-end.js", out: "session-end" },
   { entry: "dist/src/hooks/plugin-cache-gc.js", out: "plugin-cache-gc" },

harnesses/claude-code/hooks/hooks.json

@@ -31,11 +31,6 @@
             "command": "node \"${CLAUDE_PLUGIN_ROOT}/bundle/capture.js\"",
             "timeout": 10,
             "async": true
-          },
-          {
-            "type": "command",
-            "command": "node \"${CLAUDE_PLUGIN_ROOT}/bundle/recall.js\"",
-            "timeout": 2
           }
         ]
       }

src/deeplake-api.ts

@@ -290,7 +290,7 @@ export class DeeplakeApi {
   private async _queryWithRetry(sql: string, externalSignal?: AbortSignal): Promise<Record<string, unknown>[]> {
     let lastError: Error | undefined;
     for (let attempt = 0; attempt <= MAX_RETRIES; attempt++) {
-      // A caller-supplied signal (e.g. recall's latency budget) aborts the
+      // A caller-supplied signal (e.g. a hook's latency budget) aborts the
       // whole operation — including between retries — so in-flight work is
       // actually cancelled, not just abandoned.
       if (externalSignal?.aborted) throw new Error("Query aborted");

src/dir-config.ts

@@ -107,7 +107,7 @@ export interface ResolvedDirConfig {
  *
  * The two concerns are INDEPENDENT:
  *   - `orgId` / `workspaceId` are IDENTITY — they apply to reads (memory
- *     search, recall, the VFS) as well as capture. Omitted fields fall back to
+ *     search, the VFS) as well as capture. Omitted fields fall back to
  *     the global identity in `base`.
  *   - `collect` is the CAPTURE switch — writes only. It never suppresses the
  *     identity overlay, so `{ "collect": false, "workspaceId": "x" }` reads
@@ -156,7 +156,7 @@ export function resolveDirConfig(
  * THE single entry point for a workspace-scoped Config.
  *
  * Any code path that builds a `DeeplakeApi` against per-directory workspace data
- * — CLI commands (goals, rules, skills), memory read/write hooks, recall — MUST
+ * — CLI commands (goals, rules, skills), memory read/write hooks — MUST
  * get its config from here, never from a bare `loadConfig()`. It folds the
  * nearest `.hivemind` (and the `HIVEMIND_*` env locks, via resolveDirConfig)
  * into one place, so routing can never again be half-wired across call sites.

src/embeddings/embed-summary.ts

@@ -1,9 +1,10 @@
 // Robust summary embedding for the finalize (wiki-worker) path.
 //
 // Background: ~75% of production summary rows have a NULL summary_embedding,
-// so proactive recall silently degrades to weak lexical matching for most of
-// the corpus. A large share of those NULLs are NOT "embeddings disabled" — they
-// are ENABLED users whose embed daemon was cold at the one moment finalize ran.
+// so semantic memory search silently degrades to weak lexical matching for
+// most of the corpus. A large share of those NULLs are NOT "embeddings
+// disabled" — they are ENABLED users whose embed daemon was cold at the one
+// moment finalize ran.
 //
 // EmbedClient.embed() is built for the latency-critical capture hook: on a cold
 // daemon it returns null IMMEDIATELY while spawning the daemon in the