Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.jitera.ai/llms.txt

Use this file to discover all available pages before exploring further.

All data stores, caches, and message brokers are bundled in the Helm chart and deployed in-cluster by default. No external setup is required for initial deployment. For production high-availability deployments, you can externalize these services to managed alternatives (e.g., Amazon RDS, Azure Database for PostgreSQL, ElastiCache, Azure Cache for Redis). For component details, see Architecture Overview.
Jitera uses two separate PostgreSQL instances for different services:
DatabaseServicePurposeVersion
PostgreSQLAutomation (Rails)Users, projects, permissions, business logic14.x
PGVectorUltron (Node.js)AI embeddings, vector similarity search16.x
These are separate instances with different major versions and must be provisioned independently.

​
PostgreSQL (Automation)

Primary relational database used by the Automation service (Ruby on Rails) for users, projects, permissions, and business logic.
AttributeValue
Used byAutomation (Rails), Hasura
Validated version14
Supported versions14.x (all patch versions up to latest)
Required Extensions:
  • btree_gist — GiST index operator classes
  • citext — case-insensitive character string type
  • cube — multi-dimensional cube data type
  • pg_stat_statements — query execution statistics
  • pg_trgm — trigram text search
  • pgcrypto — cryptographic functions
  • uuid-ossp — UUID generation
  • vector — vector similarity search (pgvector)
These extensions are installed automatically during deployment. If you use a managed service, ensure that installation of these extensions is permitted. For provider-specific steps, see AWS EKS or Azure AKS.
Do NOT pre-create the database. The Helm migration hook creates it via db:create. Pre-creating it causes db:create to fail, which silently aborts all subsequent migrations (db:migrate), leaving the schema incomplete.

​
PGVector (Ultron)

Separate PostgreSQL instance used by the Ultron service (Node.js/NestJS) for AI embeddings and vector similarity search.
AttributeValue
Used byUltron
Validated versionPostgreSQL 16 with pgvector
Supported versions16.x (all patch versions up to latest)
Required Extensions:
  • btree_gist — GiST index operator classes
  • citext — case-insensitive character string type
  • cube — multi-dimensional cube data type
  • pg_stat_statements — query execution statistics
  • pg_trgm — trigram text search
  • pgcrypto — cryptographic functions
  • uuid-ossp — UUID generation
  • vector — vector similarity search (pgvector)
These extensions are installed automatically during deployment. If you use a managed service, ensure that installation of these extensions is permitted. For provider-specific steps, see AWS EKS or Azure AKS.
The database MUST be pre-created on the PGVector instance before running helm install. The Helm migration hook only creates the PostgreSQL (Automation) database — it does not create databases on the PGVector (Ultron) instance.
TLS limitation: External PGVector with TLS enforcement is not supported by the current Helm chart. You must disable forced TLS on the managed database instance:
  • AWS RDS: Set the rds.force_ssl parameter to 0
  • Azure Database: Set require_secure_transport to OFF
Use sslMode: disable in the Helm values. The connection remains within the VPC/VNet private subnet.

​
MongoDB

Document database for generated code artifacts, design imports, and flexible schema data.
AttributeValue
Used byAutomation
Validated version5.0
Supported versions5.0.x
The mongodb_uri must include the database name in the URI path (e.g., /jitera). If omitted, the connection defaults to the test database (MongoDB) or admin database (DocumentDB), causing operations such as index creation and data writes to target the wrong database. See MongoDB Connection String — defaultauthdb.
TLS limitation (Amazon DocumentDB): TLS must be disabled on the DocumentDB cluster when using external MongoDB. The Helm chart’s migration job cannot verify DocumentDB’s TLS certificate chain, causing migrations to silently fail. Create a cluster parameter group with tls=disabled and do not include tls=true in the mongodb_uri. The connection remains within the VPC private subnet.

​
Redis

In-memory data store for caching, sessions, and job queues.
AttributeValue
Used byAutomation
Validated version6.2 (standalone, no auth)
Supported versions6.2.x
The externalRedis.username field must be explicitly set, even if empty. If omitted, the chart inserts a placeholder string into the connection URI, causing a URI::InvalidURIError crash. Set username: "" for ElastiCache or Azure Cache for Redis.

​
RabbitMQ

Message broker for asynchronous message processing.
AttributeValue
Used byAutomation
Validated version3.13 (AWS Amazon MQ)
Supported versions3.13.x
The Jitera Helm chart ships with the in-cluster Bitnami RabbitMQ subchart at version 3.8.35 by default. The validated version above applies when switching to an external managed broker — AWS Amazon MQ currently offers only RabbitMQ 3.13, which is the version Jitera has validated in its self-hosted staging environment.
The externalRabbitmq.host must be the hostname only — do not include the protocol (amqps://) or port (:5671). The chart appends the port separately. Including the port in the host field causes a duplicate port in the connection URL.

Architecture Overview

Component descriptions and service dependencies

Helm Values

Complete Helm chart configuration reference

Deployment Requirements

Mandatory and optional deployment requirements