dcode) through Fireworks AI models. See the FireConnect overview for install and CLI basics.
Prerequisites
- Deep Agents Code (
dcode) installed - A Fireworks API key (
fw_...) or a Fire Pass key (fpk_...) - FireConnect v0.9.0+ (see Install)
Enable Fireworks routing
dcode after on or off if it is already running.
Default model
Deep Agents routes a single default model. The default isglm-fast-latest, written to config as fireworks:<model-id>.
What gets written
FireConnect edits~/.deepagents/config.toml:
- Sets
[models].defaulttofireworks:<model>and configures[models.providers.fireworks]with the Fireworks OpenAI-compatible base URL (https://api.fireworks.ai/inference) and a bakedapi_keyliteral (file mode0600) - Stores your Fireworks API key in the FireConnect keychain via
loginordeepagents on --api-key, then bakes it intoconfig.toml - FireConnect does not write
~/.deepagents/.state/auth.json. Use dcode’s/authfor credentials stored in that file
config.toml under ~/.fireconnect/deepagents/ before the first change. Running fireconnect deepagents off restores it byte-for-byte.
Browsing and picking models
fpk_...) show Fire Pass-supported routers only and cannot select firerouter.
CLI reference
fireconnect deepagents help for all options.
Switch models
Turn off Fireworks routing
config.toml from the backup in ~/.fireconnect/deepagents/.
Use a non-default config file
Fireworks on Microsoft Foundry
Deep Agents supports Fireworks on Microsoft Foundry (CLI:--provider azure or on --azure). See Microsoft Foundry in FireConnect and the portal setup guide.
--model (for example, FW-GLM-5.2), not a Fireworks serverless short ID.
FireConnect writes a fireworks-azure provider in config.toml with the Foundry base URL and deployment name. fireconnect model list only browses the Fireworks serverless catalog on the direct gateway path.
To switch back to the Fireworks gateway:
off and global config behavior.