continuedev/continue · warning · Error
Method not implemented.
Error message
Method not implemented.
What it means
The Cohere adapter does not support legacy (non-chat) text completions; completionNonStream is an intentional stub. Cohere's API is chat-oriented in this adapter, so /v1/completions style calls are unsupported.
Source
Thrown at packages/openai-adapters/src/apis/Cohere.ts:144
{
index: 0,
logprobs: undefined,
finish_reason: null,
delta: {
role: "assistant",
content: value.text,
},
},
],
usage: undefined,
};
}
}
}
completionNonStream(
body: CompletionCreateParamsNonStreaming,
): Promise<Completion> {
throw new Error("Method not implemented.");
}
completionStream(
body: CompletionCreateParamsStreaming,
): AsyncGenerator<Completion> {
throw new Error("Method not implemented.");
}
fimStream(
body: FimCreateParamsStreaming,
): AsyncGenerator<ChatCompletionChunk> {
throw new Error("Method not implemented.");
}
async rerank(body: RerankCreateParams): Promise<CreateRerankResponse> {
const endpoint = new URL("rerank", this.apiBase);
const response = await customFetch(this.config.requestOptions)(endpoint, {
method: "POST",
body: JSON.stringify(body),
headers: {
"Content-Type": "application/json",View on GitHub (pinned to 5522c6f44c)
Solutions
- Use chatCompletionNonStream instead for Cohere.
- Pick a provider that supports the completions API for base-model prompts.
- If you control the router, add capability flags so completions never reach Cohere.
Example fix
// before
await cohere.completionNonStream({ model: 'command-r', prompt: 'hi' });
// after
await cohere.chatCompletionNonStream({ model: 'command-r', messages: [{ role: 'user', content: 'hi' }] }); Defensive patterns
Strategy: type-guard
Type guard
const supportsCompletions = (p: API): boolean => !(p instanceof Cohere) && typeof p.completionNonStream === 'function';
Try / catch
try { await p.completionNonStream(b); } catch (e) { if (e.message === 'Method not implemented.') await p.chatCompletionNonStream(toChat(b)); else throw e; } Prevention
- Route all Cohere traffic through chatCompletion APIs.
- Add capability flags to your provider router.
When it happens
Trigger: Routing a CompletionCreateParamsNonStreaming request (model output raw, prompt field) to the Cohere adapter.
Common situations: Using a unified router that falls back to completion endpoints for base models (e.g. codestral-style usage) but pointing it at Cohere; migrating configs from an OpenAI-compatible provider.
Related errors
- Method not implemented.
- Failed to fetch channels: ${response.statusText}
- LastXCommitsDepth must be a number
- GitLab Private Token is required!
- No tool use found
AI-assisted analysis of continuedev/continue@5522c6f44c (2026-08-27).
Data as JSON: /api/errors/71563d1978bce344.
Report an issue: GitHub.