shumengya/sproutclaw
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(ai): add Xiaomi MiMo provider (#4005)

Browse Source 
* fix(ai): include minimax-cn in cross-provider-handoff matrix

* feat(ai): add Xiaomi MiMo provider

Adds Xiaomi MiMo as an openai-completions-compatible provider.

- packages/ai: register provider in types/KnownProvider, env-api-keys (XIAOMI_API_KEY), generate-models, models.generated.ts, overflow util, README, CHANGELOG
- packages/ai/test: extend stream, tokens, abort, empty, context-overflow, overflow, image-tool-result, tool-call-without-result, total-tokens, unicode-surrogate, cross-provider-handoff matrices with Xiaomi
- packages/coding-agent: default model (mimo-v2.5-pro), display name (Xiaomi MiMo), CLI env var docs, README, docs/providers.md

closes #3912

---------

Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
This commit is contained in:
Jake Jia
2026-05-02 06:46:05 +08:00
committed by GitHub
parent ddb8ed0c73
commit a44622670f
23 changed files with 319 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
packages/ai/src/utils/overflow.ts
View File

@@ -23,6 +23,9 @@ import type { AssistantMessage } from "../types.js";
* - Cerebras: "400/413 status code (no body)"
* - Mistral: "Prompt contains X tokens ... too large for model with Y maximum context length"
* - z.ai: Does NOT error, accepts overflow silently - handled via usage.input > contextWindow
* - Xiaomi MiMo: Truncates input to fill contextWindow exactly, then returns finish_reason "length"
* with output=0 (no room left to generate). Detected via stopReason "length" + zero output +
* input filling the context window.
* - Ollama: Some deployments truncate silently, others return errors like "prompt too long; exceeded max context length by X tokens"
*/
const OVERFLOW_PATTERNS = [
@@ -90,6 +93,8 @@ const NON_OVERFLOW_PATTERNS = [
* **Unreliable detection:**
* - z.ai: Sometimes accepts overflow silently (detectable via usage.input > contextWindow),
* sometimes returns rate limit errors. Pass contextWindow param to detect silent overflow.
* - Xiaomi MiMo: Truncates input to fit contextWindow then returns stopReason "length" with
* output=0. Pass contextWindow param to detect via the "filled context + zero output" signal.
* - Ollama: May truncate input silently for some setups, but may also return explicit
* overflow errors that match the patterns above. Silent truncation still cannot be
* detected here because we do not know the expected token count.
@@ -127,6 +132,16 @@ export function isContextOverflow(message: AssistantMessage, contextWindow?: num
}
}
// Case 3: Length-stop overflow (Xiaomi MiMo style) - server truncates oversized input
// to fit the context window, leaving no room for output. Returns stopReason "length"
// with output=0 and input+cacheRead filling the context window.
if (contextWindow && message.stopReason === "length" && message.usage.output === 0) {
const inputTokens = message.usage.input + message.usage.cacheRead;
if (inputTokens >= contextWindow * 0.99) {
return true;
}
}
return false;
}