HydraDB

View as Markdown

A database + tool node for HydraDB — a managed graph + memory store. It gives agents four tools to store text into HydraDB (with automatic knowledge-graph extraction), recall graph-enriched context by natural-language query, inspect the extracted relations, and read the database's shape.

What it does

HydraDB retrieves natively on its own servers — there is no embedding step, no query language, and no LLM translation in this node. Agents call the tools on demand; nothing flows through data lanes (lanes: {}). The node holds an API key + a target database/collection and issues HydraDB v2 REST calls.

Tools

Tool	Inputs	Returns
store_memory	text (required), metadata (optional)	ingest status + raw response
recall_memory	query (required), max_results (optional, ≤100)	ranked, graph-enriched results
query_graph	source_id (optional)	entities + relationship triplets
get_schema	—	database, collections, and infra readiness flags

Setup

1. Create an API key and a database in the HydraDB dashboard.
2. Configure the node:
   • API Key — paste it, or leave blank to use the HYDRA_DB_API_KEY environment variable.
   • Database — your HydraDB database id (required).
   • Collection — optional scope within the database (default default).
   • Max results — default cap for recall_memory (default 10).

The API key is a secure field (never stored as a default) and is cleared from memory on teardown.

Limits

• Not a vector/RAG store: there is no embedding → store → retrieve lane path — retrieval is HydraDB-native and happens via the agent tools.
• query_graph is relation inspection, not a query language — HydraDB has no Cypher/AQL equivalent; you get extracted entities and triplets, optionally scoped to one source_id.
• get_schema reports collections + infra readiness, not a per-field data schema.
• recall_memory caps max_results at 100.
• v2 scopes by database + collection (the older tenant_id / sub_tenant_id names are accepted as deprecated aliases).

Examples

Agent tool calls:

// store
{ "tool": "store_memory", "args": { "text": "Alex prefers concise answers and dark mode." } }
// recall
{ "tool": "recall_memory", "args": { "query": "What are Alex's UI preferences?", "max_results": 5 } }
// inspect the graph for one source
{ "tool": "query_graph", "args": { "source_id": "doc_12345" } }
// discover collections + readiness
{ "tool": "get_schema", "args": {} }

Upstream docs

• API reference: https://docs.hydradb.com/api-reference/v2/sdks.md
• Query: https://docs.hydradb.com/api-reference/v2/endpoint/query.md
• Ingest: https://docs.hydradb.com/api-reference/v2/endpoint/ingest-context.md
• Context relations: https://docs.hydradb.com/api-reference/v2/endpoint/source-relations.md

Troubleshooting

• "API key is required" — set the API Key field or HYDRA_DB_API_KEY.
• DATABASE_NOT_FOUND — the database id is wrong or not yet created.
• Empty recall_memory results — the database may still be provisioning; call get_schema and check infra.ready_for_ingestion.
• Requests fail with 401 — the key is invalid or lacks access to the database.

Schema

Field	Type	Description	Default
hydradb.api_key	string	API Key HydraDB API key. Falls back to the HYDRA_DB_API_KEY environment variable if left blank.
hydradb.collection	string	Collection Collection scope within the database (per-user/workspace partition for memories).	"default"
hydradb.database	string	Database HydraDB database - the space that stores your context. Create one in the HydraDB dashboard.
hydradb.max_results	integer	Max results Default maximum number of results returned by recall_memory.	10
hydradb.profile	string		"default"

Dependencies

• requests