feat: Mistral AI LLMプラグインを追加

[oboroge0]

Summary

• Mistral AIのAPIを使用するLLMプラグインを新規追加
• OpenAI互換APIを利用し、Mistral Large/Medium/Small、Codestral、Ministralモデルに対応
• グローバル設定画面にMistral AIセクション（APIキー・モデル選択）を追加
• デモモード対応（APIキー未設定時は定型文応答）

変更内容

• plugins/mistral-llm/ — プラグイン本体（manifest.json + node.ts）
• apps/server-ts/src/engine/executor.ts — グローバル設定マッピング追加
• apps/web/components/ui/SettingsModal.tsx — 設定画面にMistralセクション追加
• apps/web/components/panels/NodeSettings.tsx — ノード設定パネル対応
• apps/web/lib/constants.ts — Mistralモデルリスト追加
• apps/web/locales/{en,ja}.json — ローカライズ追加

Test plan

• npm test — 全116テスト通過
• npm run lint — 新規エラーなし
• エディタでMistral LLMノードを追加し、設定画面が正しく表示されることを確認
• グローバル設定でMistral APIキー・モデルを設定し、ノードに反映されることを確認
• APIキー設定時にMistral APIからレスポンスが返ることを確認

closes #39 (Mistral部分)

🤖 Generated with Claude Code

Summary by CodeRabbit

Release Notes

New Features

• Added Mistral AI as a new LLM provider with full integration
• Users can now configure Mistral with API key input and model selection
• Six Mistral models available: Large, Medium, Small, Codestral, Ministral 8B/3B
• Configure system prompt, temperature, and max tokens for Mistral LLM nodes
• Localization support added for English and Japanese interfaces

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR adds Mistral AI as a new LLM provider to the system. Changes include backend configuration mappings, frontend UI components for settings, model option definitions, localization entries for English and Japanese, and a new Mistral LLM plugin with manifest and node implementation following existing provider patterns.

Changes

Cohort / File(s)	Summary
Backend Configuration apps/server-ts/src/engine/executor.ts	Added "mistral-llm" mapping in GLOBAL_SETTINGS_MAP linking apiKey and model config fields to global mistral settings.
Frontend Node Settings apps/web/components/panels/NodeSettings.tsx	Extended nodeConfigs with "mistral-llm" entry including fields for API key (password), model (select dropdown), system prompt (textarea), prompt builder, and temperature (number). Updated GLOBAL_SETTINGS_MAP with mistral-llm mapping.
Settings UI Panel apps/web/components/ui/SettingsModal.tsx	Added Mistral AI configuration panel with API key input and model selection dropdowns. Added spacing adjustment (mb-4) to Ollama panel.
Model Definitions apps/web/lib/constants.ts	Added mistral key to LLM_MODEL_OPTIONS with six model options: Mistral Large, Medium, Small, Codestral, Ministral 8B, and Ministral 3B.
Localizations apps/web/locales/en.json, apps/web/locales/ja.json	Added mistral-llm node descriptions in English and Japanese locales describing the node functionality and usage requirements.
Plugin Manifest plugins/mistral-llm/manifest.json	Defined complete plugin manifest with metadata, UI configuration (label, icon, color), node I/O specification (prompt input, response output), event emissions (response.generated), and config schema with apiKey, model, systemPrompt, promptSections, temperature, and maxTokens fields.
Plugin Implementation plugins/mistral-llm/node.ts	Implemented MistralLLMNode class extending BaseNode with setup/execute/teardown lifecycle. Handles OpenAI-compatible Mistral API client initialization, demo-mode fallback, prompt construction from sections, system prompt augmentation with character context, API invocation, response emission, and comprehensive error handling.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• PR #117: Mistral LLM node integrates with shared SDK LLM error handler (handleLLMError) and follows plugin initialization patterns from this release PR.
• PR #93: Both PRs modify NodeSettings.tsx integration and plugin manifest handling for dynamic field derivation from plugin configurations.
• PR #95: Main PR extends GLOBAL_SETTINGS_MAP and model constants introduced in that PR with new Mistral entries.

Poem

🐰 A whisper through the fields of code,
Mistral winds now find their road,
With prompts and keys, a node takes flight,
Six models strong, the LLM's delight!
The config spreads from east to west,
New patterns prove the system's best.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main change: adding a Mistral AI LLM plugin. It is concise and specific about the primary feature introduced.
Linked Issues check	✅ Passed	The PR successfully implements Mistral AI support as outlined in issue #39: adds manifest.json, node.ts implementation, model selection in UI, global settings support, and localization [#39].
Out of Scope Changes check	✅ Passed	All changes are within scope of adding Mistral AI support. No unrelated modifications or out-of-scope changes detected.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch feat/mistral-llm

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (2)

apps/web/locales/ja.json (1)

103-103: Node description looks good, but nodeConfig.mistralLlm section may be missing.

The mistral-llm description correctly follows the pattern of other LLM nodes. However, other LLM providers (e.g., openaiLlm, anthropicLlm, googleLlm, ollamaLlm) have corresponding nodeConfig.* sections (lines 180-212) for field label translations. Consider adding a mistralLlm section for consistency:

"mistralLlm": {
  "label": "LLM (Mistral)",
  "apiKey": "APIキー",
  "model": "モデル",
  "systemPrompt": "システムプロンプト",
  "temperature": "温度"
}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@apps/web/locales/ja.json` at line 103, Add a corresponding nodeConfig entry
for the Mistral LLM to mirror other providers: create a "mistralLlm" object
under nodeConfig with keys "label", "apiKey", "model", "systemPrompt", and
"temperature" (use the provided Japanese translations: "LLM (Mistral)", "APIキー",
"モデル", "システムプロンプト", "温度") so the "mistral-llm" node description has consistent
field label translations like openaiLlm/anthropicLlm/googleLlm/ollamaLlm.

plugins/mistral-llm/node.ts (1)

105-129: Consider adding a defensive check for empty choices array.

The access to response.choices[0] (line 115) could throw if the API returns an empty choices array (edge case, but possible with certain error conditions).

🛡️ Optional defensive check

-      const result = response.choices[0].message.content ?? "";
+      const result = response.choices[0]?.message?.content ?? "";

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@plugins/mistral-llm/node.ts` around lines 105 - 129, Add a defensive check
before accessing response.choices[0] in the method that calls
this.client.chat.completions.create: verify that response.choices is an array
and has length > 0; if not, log an error via context.log (or call handleLLMError
with the original response/error) and return a safe fallback (e.g., empty string
or an error response) instead of indexing into choices. Update the block that
computes result (currently using response.choices[0].message.content) to first
guard on Array.isArray(response.choices) && response.choices.length > 0 and then
extract content, otherwise handle the empty case consistently with
handleLLMError and emit/return the same shaped response object.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@apps/web/locales/ja.json`:
- Line 103: Add a corresponding nodeConfig entry for the Mistral LLM to mirror
other providers: create a "mistralLlm" object under nodeConfig with keys
"label", "apiKey", "model", "systemPrompt", and "temperature" (use the provided
Japanese translations: "LLM (Mistral)", "APIキー", "モデル", "システムプロンプト", "温度") so
the "mistral-llm" node description has consistent field label translations like
openaiLlm/anthropicLlm/googleLlm/ollamaLlm.

In `@plugins/mistral-llm/node.ts`:
- Around line 105-129: Add a defensive check before accessing
response.choices[0] in the method that calls
this.client.chat.completions.create: verify that response.choices is an array
and has length > 0; if not, log an error via context.log (or call handleLLMError
with the original response/error) and return a safe fallback (e.g., empty string
or an error response) instead of indexing into choices. Update the block that
computes result (currently using response.choices[0].message.content) to first
guard on Array.isArray(response.choices) && response.choices.length > 0 and then
extract content, otherwise handle the empty case consistently with
handleLLMError and emit/return the same shaped response object.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 67bfd94a-d74d-4c6d-bfff-254c7d708ecc

📥 Commits

Reviewing files that changed from the base of the PR and between b45d01e and ddfdf10.

📒 Files selected for processing (8)

• apps/server-ts/src/engine/executor.ts
• apps/web/components/panels/NodeSettings.tsx
• apps/web/components/ui/SettingsModal.tsx
• apps/web/lib/constants.ts
• apps/web/locales/en.json
• apps/web/locales/ja.json
• plugins/mistral-llm/manifest.json
• plugins/mistral-llm/node.ts