name: slm-recall description: Search and retrieve facts, decisions, and past context from SuperLocalMemory. Use when the user asks to recall, find, search, or "what did we decide/say about X". Triggers multi-channel semantic retrieval with reranking; always call before storing anything new. when_to_use: |

"What did we decide about X?"
"Recall anything about Y"
"Do we have context on the Z feature?"
"Find stored information about authentication / the database / error handling"
"Search for what I said about Y"
Automatically before any non-trivial task, to surface prior context allowed-tools: recall, search, fetch, list_recent, Bash

slm-recall — Search & Retrieve Memory

Retrieve stored facts, decisions, and past context from SuperLocalMemory using multi-channel retrieval. The golden rule: recall before you remember.

When to use recall vs search vs fetch vs list_recent

| Situation | Tool | |-----------|------| | Conceptual or paraphrase query ("what did we agree on for auth?") | recall — full multi-channel retrieval + rerank | | Exact keyword match needed ("find facts containing BM25") | search — FTS5 BM25 only, lower latency | | You have a specific fact_id from a prior result | fetch — exact lookup, full detail | | Browse newest entries without a query | list_recent |

Use recall as the default. search is a fallback for zero-result recall on a known exact term. fetch is for when you already know the ID.

Recall-before-remember discipline

Before storing anything new, always call recall first. If a near-duplicate fact already exists, call update_memory(fact_id, content) to refine it rather than creating a duplicate. Duplicates degrade retrieval quality for every future session.

MCP-first workflow

1. Standard recall

recall(
  query="authentication strategy decision",
  limit=20,            # default 20; reduce to 5 for quick pre-task checks
  session_id="<sid>",  # pass the session_id returned by session_init
  fast=None,           # leave unset; see "Fast mode" below for what it controls
)

Real response shape (--json equivalent):

{
  "success": true,
  "results": [
    {
      "fact_id": "f8a2bc91",
      "content": "Decided to use JWT with 1h expiry for API auth (2026-06-10)",
      "score": 0.87,
      "confidence": 0.91,
      "trust_score": 0.84,
      "fact_type": "semantic",
      "channel_scores": {
        "semantic": 0.88,
        "bm25": 0.61,
        "temporal": 0.72,
        "hopfield": 0.55
      }
    }
  ],
  "count": 1,
  "query_type": "semantic",
  "channel_weights": {
    "semantic": 0.4,
    "bm25": 0.2,
    "temporal": 0.2,
    "hopfield": 0.2
  },
  "channel_status": {
    "semantic": "ok",
    "bm25": "ok",
    "temporal": "empty",
    "hopfield": "ok",
    "spreading_activation": "no_candidates",
    "entity_graph": "no_embedding",
    "profile": "disabled"
  },
  "incomplete_channels": [],
  "retrieval_time_ms": 134,
  "no_confident_match": false
}

Read channel_status before concluding that nothing is stored. It reports what each retrieval channel did on this query. channel_weights says how much each channel counts; channel_status says whether it ran at all.

| status | meaning | |---|---| | ok | the channel ran and contributed candidates | | empty | it ran and there was genuinely nothing to return | | no_candidates | it ran but nothing survived fusion | | error | it raised — its results are missing from this answer | | timeout | it exceeded its guard — results missing | | no_embedding | the query could not be embedded, so it could not run | | disabled | switched off by configuration | | not_configured | the backing service is not set up |

semantic, bm25, temporal, hopfield and spreading_activation each search and return their own candidates. profile is a shortcut that runs before them and can answer directly. entity_graph produces nothing of its own — it re-scores what the others found, by how well each result connects to the entities in your question, which is why it reports no_candidates when the rest come back empty.

empty, no_candidates, disabled and not_configured are normal. error, timeout and no_embedding mean the answer is incomplete, not negative — say so to the user rather than reporting "no memories found". incomplete_channels carries the same warning as a plain list.

Refine on low confidence. recall returns confidence signals with every result. If no_confident_match is true (or answer_confidence is low / abstained is true), do NOT invent a memory — rewrite the query into 1–3 more specific sub-queries (split multi-hop questions; try entity names, synonyms, or broader phrasing) and call recall again before concluding nothing was found. A confident match → use it directly. SLM returns fast local results (~1–2s, no server-side LLM round on the hot path) and lets you, the calling model, drive this refinement.

2. Passing session_id

Pass the session_id returned by session_init, on every recall in that session. It does two things.

It carries the conversation forward. Each recall offers its five best-ranked results to a small per-session working set of seven slots. A memory that keeps coming back is reinforced rather than duplicated, and the least-activated slot is the one evicted, so something referenced across several turns is hard to lose. Later recalls in the same session rank the held memories higher, and turn three is not as cold as turn one. The bias is deliberately small — it nudges the order, it never overrides an exact match.
It attributes engagement to the session, so a later report_outcome can close the loop on the right recall.

Omitting it costs both: recall still returns correct results, but every turn starts cold and no feedback is attributable.

Use the real id, not a made-up one. An id beginning http:, mcp:, cli: or probe: is treated as a synthetic per-request label, not a conversation, and is excluded from the working set — inventing one per call would otherwise fill the registry and evict genuine conversations.

3. Fast mode

fast controls one thing: whether the server runs its own internal LLM reformulation round. It does not disable any retrieval channel — every channel and the reranker run either way. There are four always: meaning, keyword, entity graph and time. Spreading activation and Hopfield register as a fifth and sixth when their prerequisites are present, so a store sees up to six.

Leave it unset. Unset resolves to "skip the internal round", because you are the reasoner: you refine the query yourself using the confidence signals above, and you do it better than a local model would. Pass fast=False only when SLM is deployed with no capable client in front of it.

recall(query="rate limiting approach", limit=5, session_id="<sid>")

4. Keyword fallback via search

When recall returns zero results on a specific term, try search:

search(query="BM25 indexing", limit=10, profile_id="")

profile_id="" uses the active profile. Response has success, results, and count but no channel_scores or query_type.

5. Pull full detail for a known fact

fetch(fact_ids="f8a2bc91,d4c1e203")

Returns the full record for each ID: entities, lifecycle, access_count, importance, observation_date, referenced_date. Use this when the recall summary (120-char truncation in list_recent) is not enough.

6. Browse recent memories

list_recent(limit=20, profile_id="")

Returns facts newest-first. Content is truncated to 120 chars. Use fetch once you have the fact_id for full content.

7. Close the loop — say which memories helped

Retrieval ranks a memory partly on whether it has actually been useful before. That evidence only exists if you supply it.

report_outcome(
  memory_ids="f8a2bc91,c31d0f77",   # the ids you actually used
  outcome="success",                # "success" | "failure" | "partial"
  context="used the JWT expiry decision to write the refresh handler",
)

Call it when a recall visibly changed what you did: you applied the decision, followed the convention, or avoided the gotcha. Report failure when a confidently-returned memory turned out to be wrong or stale — a negative signal is worth as much as a positive one, and it is the only way a stale memory stops being promoted.

Report only ids you genuinely used. Reporting every returned id marks the irrelevant ones useful and trains the ranker toward noise.

report_feedback(fact_id, feedback, query) is the finer-grained form for a single fact and the query that surfaced it.

How multi-channel retrieval works

recall runs multiple candidate producers in parallel — semantic vector similarity, keyword matching, temporal recency weighting, and contextual graph channels — then fuses and reranks the combined results, with an optional entity-graph score enhancement. The channel_weights field in the response shows how each channel contributed for that query. Weights adapt over time based on engagement signals attributed via session_id.

To inspect per-channel scores for a real query against your own data:

slm trace "<query>" [--limit N] [--json]

No benchmark numbers are cited here; performance is workload-dependent.

CLI fallback (when MCP is unavailable)

# Multi-channel semantic recall
slm recall "<query>" [--limit N] [--fast] [--json]

# Opt into shared/global facts for one query (v3.6.15 — off by default)
slm recall "<query>" --include-global --include-shared

# Keyword/FTS5 search (alias: slm search)
slm search "<query>" [--limit N] [--json]

# Per-channel score breakdown
slm trace "<query>" [--limit N] [--json]

# Browse recent memories
slm list [--limit N] [--json]

Flags verified in source (main.py):

slm recall: --limit, --fast, --json, --include-global / --no-global, --include-shared / --no-shared
slm search: --limit, --json
slm trace: --limit, --json
slm list: --limit / -n, --json

Multi-scope (v3.6.15, opt-in): recall is shared-OFF by default — it returns only this profile's facts. Pass --include-global / --include-shared (or the MCP include_global / include_shared args) to opt in for a query, or set the defaults in your mode_a/b/c.json config. See docs/shared-memory.md.

Flags that do NOT exist (fabricated in old skills — never write these): --min-score, --format, --project, --tags on recall or search.

Never fabricate a memory

After re-querying with refined sub-queries (see Refine on low confidence above), if no_confident_match is still true or results are empty, report it plainly. Never construct a response as if a memory was found when it was not. The user trusts that what you surface came from the store.

Multi-scope retrieval (v3.6.15+, opt-in)

By default recall returns only memories in the active profile (personal scope). To also surface memories shared from other profiles, pass the scope flags:

recall(
  query="...",
  include_global=True,   # include global-scope memories (visible to all profiles)
  include_shared=True,   # include shared-scope memories (shared with this profile)
  session_id="<sid>",
)

Scope flags are off by default. Only enable them when the user explicitly asks to see shared or global facts. See slm-scope for the full sharing model.

Profile-aware retrieval (v3.8.0+)

recall always queries the currently active profile. To query a different workspace, use switch_profile (requires code, full, or power MCP profile) before recalling, then switch back. See slm-profile for workspace switching.

Related skills

slm-remember — store the decisions and facts that recall surfaces later
slm-session — session lifecycle (must call before first recall)
slm-scope — multi-scope sharing model (personal / shared / global)
slm-profile — workspace isolation and profile switching

SuperLocalMemory v4.1.11 · Qualixar · AGPL-3.0-or-later

Rappel SLM : Recherche et récupération de mémoire

Recommandé pour