Build continuity across calls

Preserve useful evidence and constraints so later Adapt-1 calls can reason with them.

Outcome

Your application records useful evidence, decisions, constraints, or interactions and makes them available to later reasoning calls.

Workflow

record -> query -> inspect -> correct or remove -> query again

Use the /api/v1/memory/* namespace for this loop. The namespace is a stable API contract; it is not a description of the whole architecture. State is one input to Core’s continuing reasoning and adaptation.

Gateways

Need	Route	Primary result
Record an interaction	POST /api/v1/memory/store	Stored record identity/status
Retrieve ranked support	POST /api/v1/memory/search	Ranked supporting evidence
Query with reasoning	POST /api/v1/memory/query	memory_context, confidence_score, optional reasoning
Inspect expanded reasoning	POST /api/v1/memory/explain	Evidence, concepts, connections, and the contract-defined conflicts and resolution response fields
Inspect current state	POST /api/v1/memory/state	Current session state
List, delete, or clear records	POST /api/v1/memory/list, POST /api/v1/memory/delete, POST /api/v1/memory/clear	State-control response

Memory lifecycle

The remaining public memory operations belong to this same continuity loop; they are not separate API products.

Need	UAT endpoint	Contract
Correct one record while retaining its ID	POST https://rei-neuroadapt-api.reilabs.org/api/v1/memory/correct	Supply the returned memory_id plus the replacement content.
Store 1–50 records	POST https://rei-neuroadapt-api.reilabs.org/api/v1/memory/batch_store	Synchronous, input-ordered, and non-transactional.
Use the compatibility alias	POST https://rei-neuroadapt-api.reilabs.org/api/v1/batch_store	Alias for the canonical batch-store route.
Read a capacity summary	POST https://rei-neuroadapt-api.reilabs.org/api/v1/memory/state/summary	Use /memory/state with summary_only: true when full diagnostics are otherwise needed.
Preview or apply pruning	POST https://rei-neuroadapt-api.reilabs.org/api/v1/memory/prune	Start with dry_run: true; unique memories are protected by default.

A batch can partially succeed: items completed before a later failure remain stored. Inspect the per-item results before retrying. Do not issue a dependent query until the batch request returns successfully.

curl -fsS -X POST https://rei-neuroadapt-api.reilabs.org/api/v1/memory/correct \
  -H "Authorization: Bearer <api-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "session_id": "<session-id>",
    "memory_id": "<memory-id>",
    "user_message": "Corrected observation text.",
    "ai_message": "",
    "context": {"source": "operator-correction"}
  }'

Memory query, search, and explain default to update_memory_state: true; when enabled, these operations can change how evidence is organized and what remains influential in later reasoning. Set update_memory_state: false for a read-only comparison or reproducible evaluation.

Use max_records, optional max_bytes, and include_unique deliberately when pruning. Review a dry-run candidate set before repeating the call with dry_run: false.

Validation

Record one constraint, query for the decision it affects, then verify that the returned context or reasoning uses that constraint. Correct or remove the record and query again. The meaningful behavior is not storage alone; it is whether later reasoning changes appropriately.

Boundaries

• Use Domains when the application benefits from explicit entities, relations, hypotheses, rules, or domain-shaped evidence.
• Use the interval policy when the documented interval and scalar contract matches the required result.

The Quickstart contains the complete first-call example.