Configuring AIDB#

To use AIDB, add it to shared_preload_libraries , create the extensions in the database, manage user access, and verify the installation. This page also covers proxy settings and the AIDB configuration parameters (GUCs) you can tune afterward.

Configure shared_preload_libraries#

In the postgresql.conf file, add aidb to the shared_preload_libraries parameter:

shared_preload_libraries = aidb

!!!note If shared_preload_libraries already has other extensions listed, append aidb using a comma separator. The order doesn't matter. .. ::

Restart Postgres.

Create the extensions#

Create the AIDB extension in your database:

CREATE EXTENSION aidb CASCADE;

The CASCADE option automatically installs the required vector extension if it isn’t already present.

!!!Note PGD users If you’re running EDB Postgres Distributed (PGD), run the following after creating the extension to register AIDB catalog tables in the replication set:

SELECT aidb.bdr_setup();

CREATE EXTENSION pgfs;

For more information, see PGFS documentation .

Manage user access with the aidb_users role#

When you create the AIDB extension, Postgres automatically creates a role named aidb_users . Grant a user membership in this role to give it access to AIDB’s SQL functions. This role isn’t removed if you later drop the extension.

aidb_users is designed to simplify access management for AIDB features:

Purpose: Granting a user the aidb_users role gives it access to all AIDB functions, so it can use the extension’s capabilities.
Security: The aidb_users role doesn’t grant access to AIDB’s internal tables, so only the extension’s public API is exposed to users who hold it.
Background workers: When background workers are used, they execute as the user who defined them, or as a role specified by the user, provided that user has access to that role.

To let a user use AIDB functions, grant them the aidb_users role:

GRANT aidb_users TO your_user;

For example, create a role named alice with no privileges or role membership to start with:

CREATE ROLE alice
  LOGIN
  PASSWORD change_me;   -- replace with a real password

Then add alice to the aidb_users role:

GRANT aidb_users TO alice;

alice now inherits whatever privileges aidb_users has, for example EXECUTE on routines in the aidb schema.

Membership in aidb_users doesn’t grant access to your own data. Give alice (not aidb_users ) the privileges it needs on the tables AIDB will read or write. This example grants access to three arbitrary tables that would ordinarily be source tables used in a pipeline:

GRANT SELECT, INSERT, UPDATE, DELETE
  ON public.documents,
     public.email_text,
     public.chats
  TO alice;

If alice also needs to create objects in the public schema:

GRANT CREATE ON SCHEMA public TO alice;

!!!note Using AIDB through Pipeline Designer If you access AIDB through Hybrid Manager’s Pipeline Designer, pipeline operations run under a separate role, visual_pipeline_user , which must also be granted membership in aidb_users . See VPU and permissions for that setup.

Proxy settings#

If your environment routes outbound HTTP traffic through a proxy, set the HTTP_PROXY and HTTPS_PROXY environment variables in the Postgres environment file. On Ubuntu with community Postgres, this file is at /etc/postgresql/<version>/main/environment :

echo "HTTP_PROXY = http://<your-proxy-settings>/" | sudo tee -a /etc/postgresql/16/main/environment
echo "HTTPS_PROXY = http://<your-proxy-settings>/" | sudo tee -a /etc/postgresql/16/main/environment

Then restart Postgres for the changes to take effect:

sudo systemctl restart postgresql@16-main

Replace <your-proxy-settings> with your proxy address, and 16 with your Postgres version. Consult the documentation for your Postgres distribution for the exact location of the environment file.

!!!note AIDB has limited air-gapped support. The built-in CPU model runtimes (Candle and llama.cpp) run fully offline once their model files are available locally, but any pipeline step or agent tool that calls out to a remote model provider still needs outbound network access to that provider.

Configuration parameters#

AIDB exposes a set of Postgres configuration parameters (GUCs) in the aidb.* namespace. Set them in postgresql.conf , per-session with SET , or per-connection with -c . Most parameters below are USERSET — any role can change them for its own session; parameters with a different context are called out individually.

Threading limits#

Parameter	Type	Default	Range	Description
aidb.max_threads	integer	0	0–1024	Maximum number of CPU threads used for local model inference (Candle and llama.cpp). 0 uses the default: half of the available CPUs. SUSET.
aidb.max_io_threads	integer	0	0–1024	Maximum number of threads for AIDB's async I/O runtimes (remote model HTTP calls, and volume/object-store I/O). 0 falls back to the TOKIO_WORKER_THREADS environment variable, or Tokio's own default (one per CPU) if that isn't set either. SUSET.

Both apply live: a session-local SET takes effect immediately in that backend only. Setting either with ALTER SYSTEM or in postgresql.conf takes effect after pg_reload_conf() (or SIGHUP ), and propagates to all backends and background workers — no restart required.

aidb.max_threads#

SET aidb.max_threads = 4;

aidb.max_io_threads#

SET aidb.max_io_threads = 4;

See Threading limits for what these two parameters control, how they interact, and when to lower them.

Model download#

The following parameters control how AIDB downloads local model files from HuggingFace (see Local models for applicable providers).

Parameter	Type	Default	Range	Description
aidb.download_log_level	string	notice	—	Controls the output channel for routine download progress messages (for example, download start/complete, byte-progress heartbeats, and adapter load lines). Valid values: notice, log, off. See below.
aidb.download_max_attempts	integer	30	1–500	Maximum total download attempts per file (initial request plus retries) before AIDB aborts. See Local model downloads for the retry behavior.

aidb.download_log_level#

Controls where routine download-progress messages are sent. Warnings and errors are unaffected.

Value	Behavior
notice	Send messages to the client at NOTICE level. Useful for interactive sessions in psql.
log	Send messages to the server log only (subject to log_min_messages). Use this for batch jobs and tests where the client output needs to be stable across cold and warm caches.
off	Suppress routine progress messages entirely.

Example:

SET aidb.download_log_level = log;

aidb.download_max_attempts#

Caps the total number of download attempts per file. The downloader aborts early when consecutive attempts make zero progress, so this parameter is primarily a safety ceiling against excessive retry loops.

Lower it (for example, to 3 or 5 ) for fail-fast behavior in CI or scripted batch jobs. Raise it only for very large files on lossy links.

Example:

SET aidb.download_max_attempts = 5;

llama.cpp native logs#

Parameter	Type	Default	Range	Description
aidb.enable_llamacpp_logs	boolean	false	—	Controls whether llama.cpp's own diagnostic output (GGUF metadata, tensor shapes, and backend/model initialization) is written to the server log. SUSET. Read once at extension initialization — see below.

aidb.enable_llamacpp_logs#

Off by default: llama.cpp’s native logging is verbose — hundreds of lines per model load — so it’s suppressed unless you need it. This is separate from aidb.download_log_level above, which controls AIDB’s own download-progress messages; the two are independent.

Unlike the other parameters on this page, aidb.enable_llamacpp_logs is SUSET (a superuser, or a role granted SET on the parameter, can change it) and is read once when the extension initializes. Setting it with SET , ALTER SYSTEM , or editing postgresql.conf and reloading (pg_reload_conf() ) doesn’t take effect until Postgres is restarted:

aidb.enable_llamacpp_logs = true

sudo systemctl restart postgresql@16-main

Applies only to llamacpp_generate and llamacpp_embeddings models — it has no effect on Candle-based models (bert_local , clip_local , t5_local , llama_instruct_local ). See Local models — Troubleshooting for when to use this.

Network egress#

The following parameter restricts outbound HTTP traffic from AIDB — both calls to model provider APIs and model file downloads from HuggingFace. When unset, no restriction is applied and AIDB can reach any host.

Parameter	Type	Default	Description
aidb.egress_allowlist	string	unset	Comma-separated list of hosts and/or CIDR ranges that AIDB is allowed to reach. When set, any outbound request to a destination not on the list is rejected before the connection is made. SUSET.
edb.egress_allowlist	string	unset	Shared fallback allowlist used by all EDB extensions when the per-extension aidb.egress_allowlist is not set. Useful when AIDB and other EDB extensions (such as PGFS) share the same network policy. SUSET.

Entry format#

Each comma-separated entry is one of:

Form	Matches
host.example.com	Exact hostname match (case-insensitive).
.example.com	Any subdomain of example.com (doesn't match example.com itself).
203.0.113.10	Exact IPv4 or IPv6 literal in the destination URL.
203.0.113.0/24	IPv4 CIDR range; matches IP literals in that range. Doesn't match hostnames.
2001:db8::/32	IPv6 CIDR range.

Whitespace around entries is ignored. CIDR entries are checked only against IP literals — hostnames aren’t resolved for matching.

Resolution order#

For each outbound request, AIDB checks aidb.egress_allowlist first. If it isn’t set, AIDB falls back to edb.egress_allowlist . If neither is set, no check is applied. Redirects (for example, HuggingFace redirecting to a CDN host) are re-checked against the same allowlist; a redirect to a host not on the list is rejected.

Example#

Allow only direct calls to OpenAI and HuggingFace plus its CDN:

aidb.egress_allowlist = api.openai.com, huggingface.co, .huggingface.co, cdn-lfs.huggingface.co

When a request is rejected, AIDB raises an error that names the active parameter so the operator can adjust the list.

!!!note PGFS implements the same allowlist mechanism under pgfs.egress_allowlist , applied to traffic from PGFS to cloud object-store endpoints (S3, Google Cloud Storage, Azure Blob). edb.egress_allowlist is the shared fallback for both extensions: set it to apply one list to AIDB and PGFS together, or use the per-extension parameters when each needs a different policy. See PGFS — Network egress .

If your AIDB pipelines read from external storage, both allowlists are relevant: AIDB’s controls model-provider and HuggingFace traffic, and PGFS’s controls the connection to your object store. See Configuring external storage for how AIDB and PGFS work together.

aidb.create_model(credentials_env => ...) and aidb.import_mcp_tools(headers_env => ...) read a Postgres backend process environment variable, named by the caller, instead of storing a credential in the database. The following parameter restricts which variable names may be referenced this way:

Parameter	Type	Default	Description
aidb.env_var_allowed_prefix	string	AIDB_	Required prefix for environment variable names read via credentials_env/headers_env. Set to an empty string to disable the restriction. Only a superuser may change it. SUSET.

Combined with a caller-controlled provider or MCP server URL, an unrestricted variable name would let a malicious configuration read out unrelated secrets that happen to live in the Postgres process’s environment — this parameter closes that off by default while staying overridable per deployment. See Credentials from environment variables for credentials_env , and MCP tools for headers_env .

Validate the installation#

Confirm that the extensions are installed by running \dx in psql and looking for aidb , vector , pgfs (if installed), in the output:

aidb=# \dx
__OUTPUT__
                                     List of installed extensions
       Name            | Version |   Schema   |                        Description
- ----------------------+---------+------------+------------------------------------------------------------
 aidb                  | 7.6.0   | aidb       | aidb: makes it easy to build AI applications with postgres
 pgfs                  | 3.3.0   | pgfs       | pgfs: enables access to filesystem-like storage locations
 vector                | 0.8.0   | public     | vector data type and ivfflat and hnsw access methods

Configuring AIDB

Contents