# Relate cookbook

> Task-shaped recipes for Relate, the embedded openCypher graph database. Each recipe is one markdown file: runnable `cypher cell=<id>` steps, `json data=<name>` blocks holding their parameters, and after each step its recorded result as a `table` block or, for nodes and paths, a `gram` block. CI checks every result against the engine.

A step with `params=<name>` runs with that data block's JSON object as its parameters (CLI: `--param name=<json>`; Python: `db.execute(query, params)`). Each data block is also published beside its page as `data/<name>.json`.

## Agent memory

- [Agent trajectory timeline](https://relatedb.ai/cookbook/agent-trajectory-timeline/index.md): Log every tool call an agent makes as a timestamped step, then read the last N, the next page, and everything since a cutoff, with constraints that keep newest-first reads correct.
- [Bi-temporal decisions](https://relatedb.ai/cookbook/bitemporal-decisions/index.md): A decision log with two time axes answers both "what was true on this date?" and "what did the agent believe on this date?", through late corrections, a rejection, and a planned change.
- [Long-term memory for an agent](https://relatedb.ai/cookbook/agent-memory/index.md): One session writes facts and decisions from a chat into a graph; a later session recalls them, with newer decisions superseding older ones and the history kept.
- [Tool-call provenance for an agent](https://relatedb.ai/cookbook/tool-call-provenance/index.md): Record every memory tool call an agent makes in the graph it touched, then audit which call wrote a fact, who read it since, and what the failed calls left behind.

## Search and retrieval

- [Context graph for a codebase](https://relatedb.ai/cookbook/codebase-context-graph/index.md): Answer multi-hop questions about a codebase: IDF-weighted full-text search finds where to start, and bounded graph hops follow supersessions, dependencies, and citations.
- [Hybrid recall: vector + full-text + RRF](https://relatedb.ai/cookbook/hybrid-recall/index.md): Fuse vector and full-text SEARCH by reciprocal rank in one Cypher statement, then hop from each recalled note to its topic's latest decision.
- [Query the relate docs](https://relatedb.ai/cookbook/query-the-relate-docs/index.md): Find entry points in the relate docs graph with vector SEARCH, then let the graph expand them into context, scope the search by a traversal, and reach the capability and its worked example.

## Data quality

- [Entity resolution in pure Cypher](https://relatedb.ai/cookbook/entity-resolution/index.md): Resolve two messy customer exports into companies with normalized keys, bigram typo matching, IDF-weighted scores, and a human review queue, all in Cypher.
- [Rehearse a migration against a golden graph](https://relatedb.ai/cookbook/migration-rehearsal/index.md): Fingerprint a graph before a migration runs, declare what the migration may change, and catch a bug that node counts and intended-change checks both miss.

## Graph analytics

- [Dependency blast radius](https://relatedb.ai/cookbook/dependency-blast-radius/index.md): Trace every known advisory three packages can reach, measure what disappears when a dependency is dropped, and turn the answer into a CI gate.

## Geospatial

- [Nearby places for an agent](https://relatedb.ai/cookbook/nearby-places/index.md): A geo-aware memory tool that answers radius, nearest-k, map-box, and recent-visit questions over 80 remembered places.