"No API provider registered for api: undefined" with Ollama/Docker
Hey everyone, I’m hitting a wall with the latest OpenClaw build (2026.2.6-3) and could use a sanity check before I open a GitHub issue.
The Setup: I’m running a fully local stack on CachyOS (Arch) with an AMD RX 9060 XT.
Ollama: Running in Docker (ollama/ollama:rocm), serving qwen2.5-coder:14b. GPU acceleration is verified and working instantly.
OpenClaw: Running in Docker (latest), connected via --network host.
Connectivity: curl http://127.0.0.1:11434/api/tags from inside the OpenClaw container works perfectly.
The Issue: No matter how I configure the openclaw.json, the gateway crashes/throws an error as soon as I try to send a message. The error points to an issue in the internal pi-ai mapping:
What I’ve tried (and failed):
Native Ollama Provider: Configured with baseUrl pointing to localhost. models list sees it, but the gateway fails with the error above.
OpenAI Compatibility Mode: Tried using the openai provider pointing to /v1. Same error.
Config Variations: Added/removed api: "openai", api: "ollama" fields to the config. The openclaw doctor flags them as invalid input, so I removed them.
Networking: Switched from Docker Bridge IP (172.17.0.1) to Host Networking (127.0.0.1) to rule out connectivity issues.
Environment Variables: Tried bypassing the JSON config using OPENAI_BASE_URL exports.
It feels like the current build isn't correctly mapping the provider string from the config to the internal API driver, resulting in api: undefined.
Has anyone managed to get the current Docker build talking to a local Ollama instance, or is this a regression in pi-ai?
Thanks!
The Setup: I’m running a fully local stack on CachyOS (Arch) with an AMD RX 9060 XT.
Ollama: Running in Docker (ollama/ollama:rocm), serving qwen2.5-coder:14b. GPU acceleration is verified and working instantly.
OpenClaw: Running in Docker (latest), connected via --network host.
Connectivity: curl http://127.0.0.1:11434/api/tags from inside the OpenClaw container works perfectly.
The Issue: No matter how I configure the openclaw.json, the gateway crashes/throws an error as soon as I try to send a message. The error points to an issue in the internal pi-ai mapping:
What I’ve tried (and failed):
Native Ollama Provider: Configured with baseUrl pointing to localhost. models list sees it, but the gateway fails with the error above.
OpenAI Compatibility Mode: Tried using the openai provider pointing to /v1. Same error.
Config Variations: Added/removed api: "openai", api: "ollama" fields to the config. The openclaw doctor flags them as invalid input, so I removed them.
Networking: Switched from Docker Bridge IP (172.17.0.1) to Host Networking (127.0.0.1) to rule out connectivity issues.
Environment Variables: Tried bypassing the JSON config using OPENAI_BASE_URL exports.
It feels like the current build isn't correctly mapping the provider string from the config to the internal API driver, resulting in api: undefined.
Has anyone managed to get the current Docker build talking to a local Ollama instance, or is this a regression in pi-ai?
Thanks!
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
83,815Members
Resources
Similar Threads
Was this page helpful?
