Skip to main content
gligabao_lambda runs one remember, query or research in process and exits. It is for places that start for one job and are thrown away afterwards (an AWS Lambda, a container job): there is no server, HTTP API, admin UI, training, evals or daemon, and unlike the CLI it never starts a local server. The binary is static (CGO_ENABLED=0), about the size of gligabao_agent, and ./run_build.sh builds it as build/gligabao_lambda. Code: cmd/gligabao_lambda/ (main.go: flags and commands; app.go: the databases and managers).

​
How it runs

It opens the databases, runs the operation, stops the managers, closes the databases and exits:
  • remember submits the text to the remember manager and waits for the run to end, so the text, its entities, facts and embeddings are stored when the process exits. It has the pipeline of remember with its defaults.
  • query is the pipeline of query with its defaults. A user or dataset that never remembered anything gets “The knowledge base is empty: nothing was remembered yet.” (exit 0), not an error.
  • research runs a research agent and its sub-agents as gligabao_agent does (same agent definitions, tools and limits; the knowledge base is the one of the other two commands), waits for them and for the text they remembered without waiting, and prints the report. The tables of the research runs are only created for this command.
The output is the result on stdout (remember: a summary line; query: the answer; research: the report; --json: the result as JSON), and logs and totals on stderr. Exit codes: 0 ok, 1 the operation failed (a model call failed, the timeout was reached, SIGINT/SIGTERM), 2 bad flags or input, a missing API key or a database that cannot be opened. The API keys are ANTHROPIC_API_KEY and OPENAI_API_KEY (embeddings; required), from the environment or --config.

​
The state

The state is what the databases hold, and the process is thrown away, so keeping it is the caller’s job:
  • SQLite (the default): the file is --db, else $GLIGABAO_DB, else build/gligabao.db; the chromem vector store (<name>.vectors/) and the image directory (<name>.images) are next to it, or --vector-dir. The directory of the file is created if it does not exist, so on a Lambda use --db /tmp/gligabao.db. After the process exits the files are complete (the journal mode is SQLite’s default, one file with no -wal), so the caller can upload <name>.db and <name>.vectors/, and download them to the same paths before the next run.
  • --export FILE writes the whole local state to one ZIP archive when the operation ends (also if it failed), in the format of gligabao_agent --export, with gligabao_lambda as its source. It does not hold the research trajectory. This binary does not import archives: to start from a previous state, restore the SQLite file and the vector directory.
  • PostgreSQL / Pinecone (--config gligabao.toml with db_backend = "postgres" and vector_store = "pinecone"): the data persists in those servers and the process keeps nothing, so no file needs to travel. Several runs may work on one database at the same time; with SQLite files, concurrent runs from the same snapshot diverge, so run one at a time per dataset (for a Lambda: reserved concurrency of 1, or a FIFO queue).

​
Flags

Every command: --config FILE, --db FILE, --vector-dir DIR, --user (default lambda), --dataset (default main), --export FILE, --timeout (a duration like 90s; default none), --json, --verbose.
  • remember: --kind (knowledge, default, or episode), --at (the event time, RFC 3339; default now), --session.
  • query: --session.
  • research: --agent, --agent-defs-dir (default agent_defs next to the binary), --model, --max-steps (1 to 40), --max-cost-usd.
- as the argument reads it from stdin. For a Lambda, set --timeout a bit below the function’s timeout, so the databases are closed (and --export is written) before the platform kills the process.

​
Small by construction

TestMinimalDependencies in cmd/gligabao_lambda fails if the HTTP API, the server, the sidecar, training, evals, internal/db/dbfull, internal/db/dbimport, the CLI or the admin UI get linked.