chatvlm-agent1.0.2
Sign coding agents in to ChatVLM, or your own deployment of it, with one approval in the browser: an omp plugin and a login command for opencode and scripts.
Built for omp
The author tagged this package omp-plugin.
Reports 0
No reports yet.
Readme
chatvlm-agent
Use your ChatVLM models from coding agents, signed in with one approval in the browser: no key to copy, and the model list stays current.
omp
omp plugin install chatvlm-agent
omp login chatvlm # or /login chatvlm inside omp
When asked for a server, press Enter for chatvlm.com, or paste your
organization's API address (shown on its Settings → Coding agents page).
Approve the sign-in in the browser that opens, then pick a model with
/model.
opencode
opencode auth login https://api.chatvlm.com
opencode runs this package's login command, you approve in the browser, and
your models appear under ChatVLM. Use your organization's API address in
place of api.chatvlm.com on a self-hosted deployment.
Anything else
export CHATVLM_API_KEY=$(npx -y chatvlm-agent login)
prints the key, and only the key, on stdout once you approve. Point any
OpenAI-compatible tool at https://api.chatvlm.com/v1 with it.
Every sign-in makes a personal API key named after the agent and the machine. Revoke it any time on Settings → Coding agents; the agent stops working at once.
Developing
Lives in the ChatVLM monorepo (clients/chatvlm-agent). npm run build
writes dist/cli.js (the chatvlm-agent command) and dist/omp.js (the omp
plugin), each one self-contained file. The server side — the device login,
/v1, and /.well-known/opencode — is apps/chatvlm-backend/src/lib/agent-api.
Publish with npm publish from this directory; the major version is what the
server's opencode manifest pins (chatvlm-agent@1).