copilot-custom-endpoint 1.3.0 → 1.3.2

package/docs/models/kimi.md

@@ -0,0 +1,232 @@
+# Kimi — VS Code Custom Endpoint Setup Guide
+
+> **TL;DR:** Kimi K2.6 requires the local proxy. The K2 family locks `temperature: 1` and `top_p: 0.95`, and requires `thinking: { type: "disabled" }` on tool turns. The proxy rewrites sampling values, suppresses thinking on tool turns, and preserves streaming. Direct VS Code → Moonshot integration is not viable in this environment.
+
+## At a Glance
+
+| Field                  | Value                                         |
+| ---------------------- | --------------------------------------------- |
+| Mode                   | **Proxy required** (local on `:3457`)         |
+| Vision                 | ✅ Yes                                        |
+| Tool calling           | ✅ Yes (proxy forces `thinking: disabled`)    |
+| Context                | 256K                                          |
+| Max output             | 32K                                           |
+| Required `requestBody` | `temperature: 1`                              |
+| Upstream endpoint      | `https://api.moonshot.ai/v1/chat/completions` |
+| Proxy endpoint         | `http://127.0.0.1:3457/v1/chat/completions`   |
+
+## Quick Start
+
+1. **Start the proxy:** `npm run proxy:kimi`
+2. **Edit `chatLanguageModels.json`** — add the Kimi block from [Setup](#setup) below.
+3. **Set your Moonshot API key** via the Command Palette → **Chat: Manage Language Models**.
+4. **Restart VS Code** and pick "Kimi K2.6" in the chat picker.
+
+## Setup
+
+### 1. VS Code configuration
+
+Config file location:
+
+| OS      | Path                                                              |
+| ------- | ----------------------------------------------------------------- |
+| Windows | `%APPDATA%\Code\User\chatLanguageModels.json`                     |
+| macOS   | `~/Library/Application Support/Code/User/chatLanguageModels.json` |
+| Linux   | `~/.config/Code/User/chatLanguageModels.json`                     |
+
+```json
+{
+  "name": "Kimi",
+  "vendor": "customendpoint",
+  "apiKey": "",
+  "apiType": "chat-completions",
+  "models": [
+    {
+      "id": "kimi-k2.6",
+      "name": "Kimi K2.6",
+      "url": "http://127.0.0.1:3457/v1/chat/completions",
+      "requestBody": {
+        "temperature": 1
+      },
+      "toolCalling": true,
+      "vision": true,
+      "streaming": true,
+      "maxInputTokens": 262144,
+      "maxOutputTokens": 32768
+    }
+  ]
+}
+```
+
+### 2. API key
+
+1. Open the Command Palette (`Ctrl+Shift+P`).
+2. Run **Chat: Manage Language Models**.
+3. Find the **Kimi** group → **Update API Key**.
+4. Paste your Moonshot API key.
+
+> After setting via the UI, VS Code replaces `"apiKey": ""` with a `${input:chat.lm.secret.<id>}` reference.
+
+### 3. Local proxy
+
+| Setting      | Value                                                 |
+| ------------ | ----------------------------------------------------- |
+| Script       | `proxy/kimi-proxy.mjs`                                |
+| Listen URL   | `http://127.0.0.1:3457/v1/chat/completions`           |
+| Health check | `http://127.0.0.1:3457/healthz`                       |
+| Start        | `npm run proxy:kimi` (or `node proxy/kimi-proxy.mjs`) |
+| Help         | `node proxy/kimi-proxy.mjs --help`                    |
+
+#### Environment variables
+
+All can be set in a `.env` file at the repo root (both proxies `import 'dotenv/config'` automatically).
+
+| Variable                                    | Default                                               | Purpose                                                 |
+| ------------------------------------------- | ----------------------------------------------------- | ------------------------------------------------------- |
+| `KIMI_PROXY_PORT`                           | `3457` (falls back to `PORT`)                         | Local listen port                                       |
+| `KIMI_UPSTREAM_URL`                         | `https://api.moonshot.ai/v1/chat/completions`         | Upstream Moonshot endpoint                              |
+| `KIMI_PROXY_FORCE_TEMPERATURE`              | `1`                                                   | Temperature for thinking-mode requests                  |
+| `KIMI_PROXY_FORCE_NON_THINKING_TEMPERATURE` | `0.6`                                                 | Temperature when thinking is disabled (tool requests)   |
+| `KIMI_PROXY_FORCE_TOP_P`                    | `0.95`                                                | `top_p` forced into request body                        |
+| `KIMI_PROXY_DISABLE_THINKING_WITH_TOOLS`    | `1`                                                   | Force `thinking={"type":"disabled"}` when tools present |
+| `KIMI_PROXY_LOG`                            | `debug_log/kimi-proxy.ndjson` (relative to repo root) | Redacted NDJSON log path                                |
+
+#### Health check response
+
+```json
+{
+  "ok": true,
+  "upstreamUrl": "https://api.moonshot.ai/v1/chat/completions",
+  "port": 3457,
+  "forcedTemperature": 1,
+  "forcedTopP": 0.95
+}
+```
+
+#### Proxy behavior
+
+- Forwards the existing `Authorization` header upstream.
+- Rewrites plain-chat requests to `temperature: 1` and `top_p: 0.95`.
+- Rewrites tool-enabled requests to `thinking: {"type": "disabled"}`, `temperature: 0.6`, and `top_p: 0.95`.
+- Preserves streaming responses.
+- Writes redacted request summaries to `debug_log/kimi-proxy.ndjson`.
+
+## Configuration Reference
+
+### Sampling parameters
+
+| Parameter     | Value                         | Notes                            |
+| ------------- | ----------------------------- | -------------------------------- |
+| `temperature` | `1` (thinking) / `0.6` (tool) | Locked by model — proxy enforces |
+| `top_p`       | `0.95`                        | Locked by model — proxy enforces |
+
+### Thinking mode
+
+| Turn type    | Behavior                                                    |
+| ------------ | ----------------------------------------------------------- |
+| Plain chat   | Thinking enabled, `temperature: 1`                          |
+| Tool-enabled | `thinking: { type: "disabled" }` forced, `temperature: 0.6` |
+
+### Capabilities
+
+- Native multimodal: text, image, video input.
+- Tool calling with `tool_choice: "auto"`.
+- Streaming (SSE).
+- `tools` / `tool_calls` only (deprecated `functions` not supported).
+- `tool_choice="required"` is **not** supported by the model.
+
+## Troubleshooting
+
+| Symptom                                                | Likely cause               | Fix                                               |
+| ------------------------------------------------------ | -------------------------- | ------------------------------------------------- |
+| "Connection refused" on chat                           | Proxy not running          | `npm run proxy:kimi`                              |
+| `invalid temperature: only 1 is allowed`               | Direct path without proxy  | Use the proxy                                     |
+| `invalid top_p: only 0.95 is allowed`                  | Direct path without proxy  | Use the proxy                                     |
+| `thinking is enabled but reasoning_content is missing` | Tool turn with thinking on | Verify `KIMI_PROXY_DISABLE_THINKING_WITH_TOOLS=1` |
+| Model not in VS Code picker                            | Config not reloaded        | Restart VS Code                                   |
+| `tool_choice=required` rejected                        | Model limitation           | Use `auto` only                                   |
+
+## Pricing
+
+For the cross-provider comparison, see [docs/pricing.md](../pricing.md). Kimi K2.6 on the **Moonshot direct platform**:
+
+| Model       | Input      | Output (non-thinking) | Output (thinking) |
+| ----------- | ---------- | --------------------- | ----------------- |
+| `kimi-k2.6` | $0.16 / 1M | $0.95 / 1M            | $4.00 / 1M        |
+
+> Via DashScope, K2.6 is also available at $0.89 / 1M input and $3.71 / 1M output (same model, regional pricing).
+
+---
+
+## Background & Findings
+
+> This appendix preserves the validation narrative for future reference. It is not required to use the model.
+
+### Why Kimi was a reasonable candidate
+
+Kimi documents an OpenAI-compatible Chat Completions API with Bearer-token auth, `model` selection, streaming, and `tools` / `tool_calls` — making VS Code Custom Endpoint `chat-completions` mode the lowest-risk starting point.
+
+### Why direct integration failed
+
+Direct VS Code requests to Moonshot failed in stages:
+
+1. Initial auth failure while the config still pointed at the older `api.moonshot.cn` endpoint.
+2. `invalid temperature: only 1 is allowed for this model`.
+3. `invalid top_p: only 0.95 is allowed for this model`.
+4. After the first tool-enabled attempt, `thinking is enabled but reasoning_content is missing in assistant tool call message`.
+
+The model-level `requestBody.temperature = 1` override validated locally but was not sufficient in practice, which strongly suggests that VS Code's Custom Endpoint provider ignored or overwrote some model-specific request fields.
+
+### Important caveats from research
+
+- Kimi documents `tools` / `tool_calls`, not deprecated `functions` / `function_call`.
+- `tool_choice="required"` is not supported.
+- Thinking controls are Kimi-specific through a `thinking` object and `reasoning_content` fields.
+- VS Code BYOK/custom endpoint support does not replace GitHub-hosted features such as inline completions or semantic search.
+- K2-family models use fixed sampling values, which made request rewriting necessary when VS Code sent incompatible values.
+
+### Validation results
+
+| Check                                                   | Result                                                  |
+| ------------------------------------------------------- | ------------------------------------------------------- |
+| `GET /v1/models` against Moonshot                       | ✅ HTTP 200                                             |
+| Non-streaming chat against Moonshot                     | ✅ HTTP 200                                             |
+| Streaming chat against Moonshot                         | ✅ HTTP 200                                             |
+| Proxy-backed plain chat in VS Code                      | ✅                                                      |
+| Proxy-backed streaming in VS Code                       | ✅                                                      |
+| Proxy-backed integrated-browser tool use (post-rewrite) | ✅                                                      |
+| Direct VS Code → Moonshot (no proxy)                    | ❌ — fails on temperature / top_p / `reasoning_content` |
+
+### Tool-enabled validation details
+
+**Prompt:** "Please open kimi documentation site using vscode integrated browser"
+
+- First run: browser tool invocation succeeded, but the post-tool follow-up failed because thinking remained enabled and VS Code did not preserve `reasoning_content`.
+- Workaround: force `thinking: { "type": "disabled" }` plus `temperature: 0.6` on tool-enabled turns.
+- Rerun: both the tool turn and the follow-up model turn returned upstream `200` with `text/event-stream`.
+
+### Proxy validation notes
+
+- Redacted proxy logs confirmed `temperature 0.1 -> 1` and `top_p 1 -> 0.95` for plain-chat requests.
+- Redacted proxy logs later confirmed `thinking undefined -> disabled` and `temperature 0.1 -> 0.6` for tool-enabled requests.
+
+### Final verdict
+
+- Acceptable for plain chat: **yes** (proxy)
+- Acceptable for streaming chat: **yes** (proxy)
+- Acceptable for tool-enabled agent use: **yes**, with the local proxy workaround
+- Acceptable without a proxy: **no**
+
+## References
+
+- VS Code custom endpoint docs: `https://code.visualstudio.com/docs/copilot/customization/language-models#_add-a-custom-endpoint-model`
+- Kimi docs index: `https://platform.kimi.ai/docs/llms.txt`
+- Kimi chat completion docs: `https://platform.kimi.ai/docs/api/chat.md`
+- Kimi models list: `https://platform.kimi.ai/docs/api/list-models.md`
+- Kimi model parameter reference: `https://platform.kimi.ai/docs/api/models-overview.md`
+- Kimi tool use docs: `https://platform.kimi.ai/docs/api/tool-use.md`
+- Kimi K2.6 quickstart: `https://platform.kimi.ai/docs/guide/kimi-k2-6-quickstart.md`
+- Kimi thinking guide: `https://platform.kimi.ai/docs/guide/use-kimi-k2-thinking-model.md`
+- Kimi web search guide: `https://platform.kimi.ai/docs/guide/use-web-search.md`
+- Kimi coding tools / agent guide: `https://platform.kimi.ai/docs/guide/agent-support.md`
+- Kimi K2.6 pricing: `https://platform.kimi.ai/docs/pricing/chat-k26`

package/docs/models/mimo.md

@@ -0,0 +1,258 @@
+# Xiaomi MiMo — VS Code Custom Endpoint Setup Guide
+
+> **TL;DR:** MiMo works directly — no proxy needed. Set `thinking: { type: "disabled" }` in `requestBody` for tool-loop stability, because MiMo's API rejects (HTTP 400) any tool turn that is missing historical `reasoning_content`. Disabling thinking eliminates the field, so loops stay stable.
+
+## At a Glance
+
+| Field                  | Value                                            |
+| ---------------------- | ------------------------------------------------ |
+| Mode                   | **Direct** (no proxy)                            |
+| Vision                 | ✅ Yes (`mimo-v2.5` only)                        |
+| Tool calling           | ✅ Yes (with `thinking: disabled`)               |
+| Context                | 1M (V2.5 Pro / V2.5) / 256K (V2 Flash)           |
+| Max output             | 128K (V2.5 Pro) / 32K (V2.5) / 64K (V2 Flash)    |
+| Required `requestBody` | `thinking: { type: "disabled" }`                 |
+| Endpoint               | `https://api.xiaomimimo.com/v1/chat/completions` |
+
+### Models at a glance
+
+| Model           | Vision | Context | Role                                       |
+| --------------- | ------ | ------- | ------------------------------------------ |
+| `mimo-v2.5-pro` | ❌     | 1M      | Flagship text-only — best for agentic work |
+| `mimo-v2.5`     | ✅     | 1M      | Omnimodal — text + image + video + audio   |
+| `mimo-v2-flash` | ❌     | 256K    | Fastest and cheapest — strong reasoning    |
+
+> Legacy `mimo-v2-pro` and `mimo-v2-omni` auto-route to V2.5 (with V2.5 pricing) as of June 1, 2026, and will be fully deprecated by June 30, 2026. Use the V2.5 series.
+
+## Quick Start
+
+1. **Edit `chatLanguageModels.json`** — add the MiMo block(s) from [Setup](#setup) below.
+2. **Set your `MIMO_API_KEY`** via Command Palette → **Chat: Manage Language Models**.
+3. **Restart VS Code** and pick "MiMo V2.5 Pro", "MiMo V2.5", or "MiMo V2 Flash".
+
+## Setup
+
+### 1. VS Code configuration
+
+Config file location:
+
+| OS      | Path                                                              |
+| ------- | ----------------------------------------------------------------- |
+| Windows | `%APPDATA%\Code\User\chatLanguageModels.json`                     |
+| macOS   | `~/Library/Application Support/Code/User/chatLanguageModels.json` |
+| Linux   | `~/.config/Code/User/chatLanguageModels.json`                     |
+
+```json
+{
+  "name": "MiMo",
+  "vendor": "customendpoint",
+  "apiKey": "",
+  "apiType": "chat-completions",
+  "models": [
+    {
+      "id": "mimo-v2.5-pro",
+      "name": "MiMo V2.5 Pro",
+      "url": "https://api.xiaomimimo.com/v1/chat/completions",
+      "toolCalling": true,
+      "vision": false,
+      "streaming": true,
+      "maxInputTokens": 1048576,
+      "maxOutputTokens": 131072,
+      "requestBody": {
+        "thinking": { "type": "disabled" },
+        "temperature": 1,
+        "top_p": 0.95
+      }
+    },
+    {
+      "id": "mimo-v2.5",
+      "name": "MiMo V2.5",
+      "url": "https://api.xiaomimimo.com/v1/chat/completions",
+      "toolCalling": true,
+      "vision": true,
+      "streaming": true,
+      "maxInputTokens": 1048576,
+      "maxOutputTokens": 32768,
+      "requestBody": {
+        "thinking": { "type": "disabled" },
+        "temperature": 1,
+        "top_p": 0.95
+      }
+    },
+    {
+      "id": "mimo-v2-flash",
+      "name": "MiMo V2 Flash",
+      "url": "https://api.xiaomimimo.com/v1/chat/completions",
+      "toolCalling": true,
+      "vision": false,
+      "streaming": true,
+      "maxInputTokens": 262144,
+      "maxOutputTokens": 65536,
+      "requestBody": {
+        "thinking": { "type": "disabled" },
+        "temperature": 0.3,
+        "top_p": 0.95
+      }
+    }
+  ]
+}
+```
+
+### 2. API key
+
+1. Open the Command Palette (`Ctrl+Shift+P`).
+2. Run **Chat: Manage Language Models**.
+3. Find the **MiMo** group → **Update API Key**.
+4. Paste your MiMo API key.
+
+> After setting via the UI, VS Code replaces `"apiKey": ""` with a `${input:chat.lm.secret.<id>}` reference.
+
+### 3. Token Plan (optional)
+
+Token Plan subscribers use different base URLs and `tp-` prefixed keys:
+
+| Protocol  | Base URL                                         |
+| --------- | ------------------------------------------------ |
+| OpenAI    | `https://token-plan-cn.xiaomimimo.com/v1`        |
+| Anthropic | `https://token-plan-cn.xiaomimimo.com/anthropic` |
+
+> Pay-as-you-go keys are `sk-…`; Token Plan keys are `tp-…`. The endpoint to use depends on which key you set.
+
+## Configuration Reference
+
+### Sampling parameters
+
+| Task type            | `temperature` | `top_p` |
+| -------------------- | ------------- | ------- |
+| Agentic / tool-use   | `0.3`         | `0.95`  |
+| Vibe coding          | `0.3`         | `0.95`  |
+| General conversation | `0.8`         | `0.95`  |
+| Math reasoning       | `1.0`         | `0.95`  |
+
+> For `mimo-v2.5-pro` and `mimo-v2.5`, MiMo's docs recommend `temperature: 1.0` and `top_p: 0.95` regardless of task. In thinking mode these models also **lock** `temperature` to `1.0` — any custom value is silently overridden. Since we disable thinking, your `requestBody` value is honored.
+
+MiMo accepts `temperature` in `[0, 1.5]` and `top_p` in `[0.01, 1.0]`.
+
+### Thinking mode
+
+| Model                        | API default `thinking.type` | API default `temperature`  |
+| ---------------------------- | --------------------------- | -------------------------- |
+| `mimo-v2.5-pro`, `mimo-v2.5` | `enabled`                   | `1.0` (locked in thinking) |
+| `mimo-v2-flash`              | `disabled`                  | `0.3` (customizable)       |
+
+When thinking is enabled, responses include a `reasoning_content` field alongside `content` and `tool_calls`.
+
+### Capabilities
+
+- Streaming (SSE, standard OpenAI format).
+- Tool calling with `tool_choice: "auto"`.
+- Vision (image input via OpenAI `content` array) on `mimo-v2.5` only.
+- `tool_choice` other than `"auto"` is **stripped** and treated as `"auto"`.
+- `mimo-v2.5` also supports video and audio understanding.
+
+### Rate limits
+
+**100 RPM / 10M TPM** per model per account.
+
+## Troubleshooting
+
+| Symptom                                    | Likely cause                                             | Fix                                                   |
+| ------------------------------------------ | -------------------------------------------------------- | ----------------------------------------------------- |
+| HTTP 400 on the second turn of a tool loop | `reasoning_content` missing in history (thinking on)     | Add `thinking: { type: "disabled" }` to `requestBody` |
+| Vision request returns an error            | Used `mimo-v2.5-pro` or `mimo-v2-flash` (text-only)      | Use `mimo-v2.5` for vision                            |
+| Custom `tool_choice` ignored               | MiMo only honors `"auto"`                                | Stick to `auto`                                       |
+| 401 Unauthorized                           | Wrong key, or Token Plan URL used with pay-as-you-go key | Match key prefix (`sk-` vs `tp-`) to the endpoint     |
+| 429 rate-limited                           | Concurrent sessions exceeded 100 RPM / 10M TPM           | Reduce concurrent agent sessions                      |
+
+## Pricing
+
+For the cross-provider comparison, see [docs/pricing.md](../pricing.md). Overseas (international) pay-as-you-go rates:
+
+| Model           | Input (Cache Hit) | Input (Cache Miss) | Output     |
+| --------------- | ----------------- | ------------------ | ---------- |
+| `mimo-v2.5-pro` | $0.20 / 1M        | $1.00 / 1M         | $3.00 / 1M |
+| `mimo-v2.5`     | $0.08 / 1M        | $0.40 / 1M         | $2.00 / 1M |
+| `mimo-v2-flash` | $0.01 / 1M        | $0.10 / 1M         | $0.30 / 1M |
+
+> Cache writing is currently free of charge (limited-time offer). MiMo also offers a Token Plan subscription with discounted rates and a free cache-writing promotion.
+
+---
+
+## Background & Findings
+
+> This appendix preserves the validation narrative for future reference. It is not required to use the model.
+
+### The critical `reasoning_content` constraint
+
+When thinking mode is enabled and the conversation history contains tool calls, the `reasoning_content` field **must** be fully passed back in every subsequent assistant message. Otherwise, the API returns HTTP 400.
+
+This is the same class of problem as Qwen's `reasoning_content` issue, but **stricter**: MiMo's API actively rejects requests with missing historical `reasoning_content`, rather than silently degrading.
+
+**Implication for VS Code Copilot:** VS Code's agent mode is unlikely to preserve `reasoning_content` across multi-turn tool loops. Therefore:
+
+- **Thinking enabled + tool calling = broken** (400 errors after the first tool round-trip).
+- **Thinking disabled + tool calling = works** (no `reasoning_content` to preserve).
+- **Thinking enabled + plain chat = works** (no tool calls in history).
+
+### Why a static `thinking: disabled` is enough
+
+VS Code's agent mode is the only flow that triggers tool loops, and we already disable thinking for those turns. Plain chat with thinking enabled works fine because no `reasoning_content` accumulates in history.
+
+A dynamic proxy (suppress thinking only when tools are present — same pattern as `proxy/qwen-proxy.mjs`) would let plain chat show reasoning, but it is **not implemented** because:
+
+- The cost of losing visible reasoning in plain chat is low for most users.
+- Static suppression is one less moving part to maintain.
+
+### Benchmark highlights (from official MiMo V2.5 announcement)
+
+| Model           | SWE-Bench Verified | SWE-Bench Pro | Terminal-Bench 2.0 | AIME 2025 |
+| --------------- | ------------------ | ------------- | ------------------ | --------- |
+| `mimo-v2.5-pro` | —                  | 57.2%         | 68.4%              | —         |
+| `mimo-v2.5`     | —                  | 56.1%         | —                  | —         |
+| `mimo-v2-flash` | 73.4%              | —             | —                  | 94.1%     |
+
+> `mimo-v2.5` additionally scores 87.7% on Video-MME and 62.3% on Claw-Eval Text.
+
+### Validation results
+
+| #   | Test                                      | Model           | Result                                                                                |
+| --- | ----------------------------------------- | --------------- | ------------------------------------------------------------------------------------- |
+| 1   | Add provider to `chatLanguageModels.json` | All             | ✅                                                                                    |
+| 2   | Plain chat in VS Code                     | `mimo-v2.5-pro` | ✅ — model self-identified as MiMo 1T-param                                           |
+| 3   | Agent mode (tool calling)                 | `mimo-v2.5-pro` | ✅ — file reads, browser automation, terminal, image viewing all worked               |
+| 4   | Vision                                    | `mimo-v2.5`     | ✅ — analyzed an attached screenshot (Facebook post, browser tabs, sidebar) in detail |
+
+External API checks (curl):
+
+| Check              | Model           | Result                                                  |
+| ------------------ | --------------- | ------------------------------------------------------- |
+| Non-streaming chat | `mimo-v2-flash` | ✅                                                      |
+| Streaming (SSE)    | `mimo-v2-flash` | ✅                                                      |
+| Non-streaming chat | `mimo-v2.5-pro` | ✅                                                      |
+| Tool calling       | `mimo-v2-flash` | ✅ — `finish_reason: "tool_calls"` with valid JSON args |
+
+### Known risks
+
+| Risk                                  | Detail                                                             | Mitigation                                              |
+| ------------------------------------- | ------------------------------------------------------------------ | ------------------------------------------------------- |
+| `reasoning_content` 400 errors        | If thinking is accidentally enabled in tool loops, API returns 400 | Always set `thinking.type: "disabled"` in `requestBody` |
+| `tool_choice` only supports `"auto"`  | Non-`auto` values are stripped                                     | Should not affect VS Code, which uses `auto`            |
+| Auth header format                    | Both `api-key:` and `Authorization: Bearer` work                   | VS Code sends `Authorization: Bearer` — works directly  |
+| `temperature` locked in thinking mode | V2.5 Pro / V2.5 force `temperature: 1.0` when thinking is on       | Not an issue when thinking is disabled                  |
+| 1M context window                     | VS Code may not send enough tokens to benefit                      | Set conservatively; adjust after testing                |
+
+## References
+
+- API Platform: `https://platform.xiaomimimo.com/`
+- OpenAI API Reference: `https://platform.xiaomimimo.com/docs/en-US/api/chat/openai-api`
+- First API Call Guide: `https://platform.xiaomimimo.com/docs/en-US/quick-start/first-api-call`
+- Model & Rate Limits: `https://platform.xiaomimimo.com/docs/en-US/quick-start/model`
+- Model Hyperparameters: `https://platform.xiaomimimo.com/docs/en-US/quick-start/model-hyperparameters`
+- Pricing: `https://platform.xiaomimimo.com/docs/en-US/pricing`
+- `reasoning_content` Guide: `https://platform.xiaomimimo.com/docs/en-US/usage-guide/passing-back-reasoning_content`
+- AI Tools Integration: `https://platform.xiaomimimo.com/docs/en-US/integration/claude-code`
+- HuggingFace (MiMo-V2.5): `https://huggingface.co/XiaomiMiMo/MiMo-V2.5`
+- HuggingFace (MiMo-V2.5-Pro): `https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro`
+- HuggingFace (MiMo-V2-Flash): `https://huggingface.co/XiaomiMiMo/MiMo-V2-Flash`
+- MiMo V2.5 Blog: `https://mimo.xiaomi.com/mimo-v2-5`
+- AI Studio (playground): `https://aistudio.xiaomimimo.com/`