AIMO Docs

Agent source ​

Everything in the browser below is mirrored verbatim from the agent's source tree in the AIMO repository: the six packages the published image is built from (agent, core, dbtype, executor, jobs and monitors). Nothing here is a simplified or rewritten example.

A few starting points:

  • Connection loop (agent/jobs/agent_worker.py): the outbound WebSocket connection to AIMO: backoff, JWT refresh, idle reconnect, the queue that carries results across a disconnect. See Security: network boundary.
  • Running monitors (agent/monitor/process.py): turns monitor definitions into a grouped SQL query and returns only the aggregates. See Security: aggregates, not raw rows.
  • Validation (agent/monitor/validate_artifacts.py, agent/monitor/validate_table_characteristics.py): the two safety gates of Operations, run on your database before a monitor is stored.
  • Compiling a spec (monitors/compiler/, monitors/sql_validation.py): the family algebra becoming one bounded statement, and the syntax-tree check every free-form fragment passes first.
  • Credential encryption (executor/credentials.py): scrypt and AES-256-GCM, applied before anything leaves your environment. See Security: database credentials.
  • The database adapters (dbtype/): one dialect per engine, with the read-only session mode and the regex spelling each one uses.
  • Configuration surface (agent/settings.py): the environment variables behind the Agent CLI requirements.
Browse the agent source75 files, mirrored verbatim — tree navigation, full syntax highlighting.
  • Architecture: network layout and trust boundaries.
  • Security: the properties this code implements.
  • Agent CLI: the commands built on top of this code.
  • Operations: the job types the running agent executes.