@chappibunny/repolens 0.9.0 → 1.2.0

package/CHANGELOG.md

@@ -2,6 +2,99 @@
 
 All notable changes to RepoLens will be documented in this file.
 
+## 1.2.0
+
+### ✨ Config Migration
+
+- **`repolens migrate` now patches `.repolens.yml`**: Automatically adds `configVersion: 1` to legacy config files that are missing it. Creates a `.repolens.yml.backup` before modifying. Supports both `.yml` and `.yaml` extensions, dry-run mode, and graceful handling of parse errors or missing configs.
+
+### � Bug Fixes
+
+- **GitHub Wiki empty pages**: Fixed branch mismatch where `git init` created a `main` branch but GitHub Wiki serves from `master`. Now explicitly initializes with `master` and pushes to `refs/heads/master`. Also added content validation to skip pages with empty content.
+
+### �📝 Documentation
+
+- Updated all documentation to reflect v1.2.0 version references
+- Updated TROUBLESHOOTING.md with `migrate` auto-fix hint for missing `configVersion`
+- Updated KNOWN_ISSUES.md planned improvements (removed shipped features)
+- Updated ROADMAP.md with v1.1.0 shipped items
+
+## 1.1.0
+
+### ✨ GitHub Wiki Publisher UX Enhancement
+
+Major upgrade to the wiki output quality — better information hierarchy, audience-aware navigation, and richer page structure.
+
+- **Audience-grouped Home page**: Pages organized by reader (Stakeholders, Engineers, New Contributors, Change Tracking) instead of a flat list
+- **Page descriptions**: Each link on Home includes a one-line summary of the page's purpose
+- **Status rail**: Home page now shows a metadata table (project, branch, page count, publisher, source)
+- **Recommended reading order**: Guided path through the docs for first-time readers
+- **Grouped sidebar**: Navigation split into Overview and Architecture sections (plus Custom Pages)
+- **Page metadata headers**: Every page gets a `[← Home](Home)` back-link, audience tag, and branch indicator
+- **Cleaner footer**: Compact format with branch context and Home link
+- **New constants**: `PAGE_DESCRIPTIONS`, `AUDIENCE_GROUPS`, `SIDEBAR_GROUPS`, `PAGE_AUDIENCE`
+- **New helpers**: `getPageDisplayTitle()`, `getCustomPageKeys()`, `wikiLink()`, `pageHeader()`
+
+### 🐛 Bug Fixes
+- **Temperature still sent to GPT-5**: `DEFAULT_TEMPERATURE = 0.2` always leaked into API requests even after v1.0.1 fixes — the fallback chain (`temperature ?? aiConfig.temperature ?? DEFAULT_TEMPERATURE`) guaranteed it was never `undefined`. Removed the default entirely; temperature is now only sent when explicitly configured via `REPOLENS_AI_TEMPERATURE` env var or `ai.temperature` in config.
+
+## 1.0.1
+
+### 🐛 Bug Fixes
+- **GPT-5 API compatibility**: Use `max_completion_tokens` instead of deprecated `max_tokens` parameter
+- **GPT-5 temperature handling**: Omit `temperature` from API requests for models that only support the default value (e.g. gpt-5-mini)
+- Removed all hardcoded `temperature: 0.2` overrides from AI section generators
+- Removed `REPOLENS_AI_TEMPERATURE` from CI workflow and `.env.example`
+- Updated `init` scaffolding to omit temperature from default config
+
+### 🔧 Improvements
+- Upgraded GitHub Actions: checkout v4.2.2, setup-node v4.2.0, Node.js 22
+
+## 1.0.0
+
+### 🎉 Stable Release
+
+RepoLens v1.0.0 marks the first stable release with a frozen public API. All CLI commands, configuration schema, and plugin interfaces are now covered by semantic versioning guarantees. See [STABILITY.md](STABILITY.md) for the full contract.
+
+### ✨ New Features
+- **GitHub Wiki Publisher**: Publish documentation directly to your repository's Wiki tab
+  - Git-based: clones wiki repo, writes pages, commits and pushes
+  - Generates `Home.md` index, `_Sidebar.md` navigation, `_Footer.md`
+  - Branch filtering via `github_wiki.branches` in `.repolens.yml`
+  - Auto-detects repository from `GITHUB_REPOSITORY` env or git remote
+  - Token sanitization in error messages (never leaks `GITHUB_TOKEN`)
+  - Config: `sidebar` and `footer` toggles
+  - 17 tests covering publishing, branch filtering, config validation, and security
+
+### 🐛 Bug Fixes
+- **GPT-5 API compatibility**: Use `max_completion_tokens` instead of deprecated `max_tokens` parameter
+- **GPT-5 temperature handling**: Omit `temperature` from API requests for models that only support the default value (e.g. gpt-5-mini)
+- **Git identity in wiki publisher**: Set committer name/email (`RepoLens Bot`) so CI runners can commit
+- **Publisher allowlist**: Added `github_wiki` to `validate.js` publisher validation (was only in `config-schema.js`)
+- **Doctor false-success**: `repolens doctor` now correctly exits with code 2 when `runDoctor()` reports failures (previously always printed "validation passed")
+- **Feedback exit code**: `repolens feedback` now exits with code 1 when feedback fails to send (previously exited 0)
+- **Unknown flags**: `repolens --unknown-flag` now prints an error instead of silently running publish
+- **scan.ignore security bypass**: Security validator now checks `scan.ignore` patterns (was incorrectly checking non-existent `scan.exclude`)
+- **Domains type mismatch**: Both validators now expect `domains` as an object (was array in security validator, object in schema validator)
+- **Plugin publisher crash isolation**: Plugin publisher errors are now caught and logged instead of crashing the pipeline
+
+### ✅ Config Stability
+- `configVersion: 1` is now **required** (was optional in schema validator)
+- Added `confluence` config section validation (branches array)
+- Added `ai.temperature` range validation (0–2)
+- Added `ai.max_tokens` range validation (>0)
+- AI config values from `.repolens.yml` are now used as fallbacks when env vars are not set
+
+### 🔧 Improvements
+- Standardized exit codes: `EXIT_SUCCESS=0`, `EXIT_ERROR=1`, `EXIT_VALIDATION=2` as named constants
+- Replaced all `console.log`/`console.warn` in production code with logger utilities
+- Removed stale "v0.4.0" references from CLI help text and migrate command
+- Sentry DSN moved to `REPOLENS_SENTRY_DSN` env var (with backwards-compatible default)
+- Hardcoded email in `init` scaffolding replaced with `your-email@example.com` placeholder
+
+### 📄 Documentation
+- Added [STABILITY.md](STABILITY.md): Complete public API contract (CLI, config schema, plugin interface, exit codes, env vars)
+
 ## 0.9.0
 
 ### ✨ New Features

package/README.md

@@ -16,11 +16,11 @@
 
 AI-assisted documentation intelligence system that generates architecture docs for engineers AND readable system docs for stakeholders
 
-**Current Status**: v0.9.0 — Plugin System
+**Current Status**: v1.2.0 — Stable Release
 
 RepoLens automatically generates and maintains living architecture documentation by analyzing your repository structure, extracting meaningful insights from your package.json, and creating visual dependency graphs. Run it once, or let it auto-update on every push.
 
-⚠️ **Early Access Notice**: The CLI commands and configuration format may evolve until v1.0. We will provide migration guides for any breaking changes. v1.0 will guarantee stable CLI behavior and config schema.
+The CLI, configuration schema, and plugin interface are **stable** as of v1.0. See [STABILITY.md](STABILITY.md) for the full API contract and semver guarantees.
 
 ---
 
@@ -49,7 +49,7 @@ For Confluence:
 ```bash
 # Edit .env and add:
 CONFLUENCE_URL=https://your-company.atlassian.net/wiki
-CONFLUENCE_EMAIL=trades@rabitaitrades.com
+CONFLUENCE_EMAIL=your-email@example.com
 CONFLUENCE_API_TOKEN=your-token
 CONFLUENCE_SPACE_KEY=DOCS
 ```
@@ -124,7 +124,7 @@ RepoLens automatically detects:
 ✅ **Deterministic Fallback** - Always generates docs even if AI unavailable  
 ✅ **Business Domain Inference** - Automatically maps code to business functions  
 ✅ **Data Flow Analysis** - Understands how information moves through your system  
-✅ **Multiple Publishers** - Output to Notion, Confluence, Markdown, or all three  
+✅ **Multiple Publishers** - Output to Notion, Confluence, GitHub Wiki, Markdown, or all four  
 ✅ **Branch-Aware** - Prevent doc conflicts across branches  
 ✅ **GitHub Actions** - Autonomous operation on every push  
 ✅ **Team Notifications** - Discord integration with rich embeds (NEW in v0.6.0)  
@@ -137,6 +137,8 @@ RepoLens automatically detects:
 ✅ **TypeScript Type Graph** - Map interfaces, classes, and type relationships (NEW in v0.8.0)  
 ✅ **Dependency Graph** - Import analysis with circular dependency detection (NEW in v0.8.0)  
 ✅ **Architecture Drift** - Track structural changes against a baseline (NEW in v0.8.0)  
+✅ **Plugin System** - Extend with custom renderers, publishers, and hooks (NEW in v0.9.0)  
+✅ **Stable API** - CLI, config, and plugin interface frozen with semver guarantees (v1.0.0)  
 
 ---
 
@@ -228,10 +230,12 @@ npm link
 Install from a specific version:
 
 ```bash
-npm install https://github.com/CHAPIBUNNY/repolens/releases/download/v0.9.0/chappibunny-repolens-0.9.0.tgz
+npm install https://github.com/CHAPIBUNNY/repolens/releases/download/v1.2.0/chappibunny-repolens-1.2.0.tgz
 ```
 </details>
 
+> **Having install problems?** See [TROUBLESHOOTING.md](TROUBLESHOOTING.md) for solutions to common issues like `ENOTEMPTY`, `npm ci` failures, and stale versions.
+
 ---
 
 ## 🎓 Complete Onboarding Guide
@@ -311,6 +315,14 @@ publishers:
 ```
 Maximum visibility: Notion for async collaboration, Confluence for enterprise docs, Markdown for local backups.
 
+**GitHub Wiki** (ideal for open source):
+```yaml
+publishers:
+  - github_wiki
+  - markdown
+```
+Docs live alongside your code — accessible from the repo's Wiki tab. Requires `GITHUB_TOKEN`.
+
 ### Step 3: Enable AI Features (Optional)
 
 **AI-enhanced documentation adds natural language explanations for non-technical audiences.**
@@ -318,7 +330,7 @@ Maximum visibility: Notion for async collaboration, Confluence for enterprise do
 **3.1: Choose an AI Provider**
 
 RepoLens works with any OpenAI-compatible API:
-- **OpenAI** (gpt-4-turbo-preview, gpt-3.5-turbo)
+- **OpenAI** (gpt-5-mini, gpt-5.4, gpt-5-nano)
 - **Anthropic Claude** (via API gateway)
 - **Azure OpenAI** (enterprise deployments)
 - **Local Models** (Ollama, LM Studio, etc.)
@@ -333,8 +345,7 @@ REPOLENS_AI_API_KEY=sk-xxxxxxxxxxxxx
 
 # Optional: Customize provider
 REPOLENS_AI_BASE_URL=https://api.openai.com/v1
-REPOLENS_AI_MODEL=gpt-4-turbo-preview
-REPOLENS_AI_TEMPERATURE=0.3
+REPOLENS_AI_MODEL=gpt-5-mini
 REPOLENS_AI_MAX_TOKENS=2000
 ```
 
@@ -344,7 +355,6 @@ REPOLENS_AI_MAX_TOKENS=2000
 ai:
   enabled: true              # Enable AI features
   mode: hybrid               # hybrid, full, or off
-  temperature: 0.3           # Lower = more focused (0.0-1.0)
   max_tokens: 2000           # Token limit per request
 
 features:
@@ -356,7 +366,7 @@ features:
   change_impact: true        # Architecture diff with context
 ```
 
-**Cost Estimates** (with gpt-4-turbo-preview):
+**Cost Estimates** (with gpt-5-mini):
 - Small repo (<50 files): $0.10-$0.30 per run
 - Medium repo (50-200 files): $0.30-$0.80 per run
 - Large repo (200+ files): $0.80-$2.00 per run
@@ -428,7 +438,7 @@ If using the Confluence publisher:
 Create `.env` in your project root:
 ```bash
 CONFLUENCE_URL=https://your-company.atlassian.net/wiki
-CONFLUENCE_EMAIL=trades@rabitaitrades.com
+CONFLUENCE_EMAIL=your-email@example.com
 CONFLUENCE_API_TOKEN=your-api-token-here
 CONFLUENCE_SPACE_KEY=DOCS
 CONFLUENCE_PARENT_PAGE_ID=123456789  # Optional
@@ -449,7 +459,7 @@ Add as repository secrets:
 2. Click **"New repository secret"**
 3. Add:
    - Name: `CONFLUENCE_URL`, Value: `https://your-company.atlassian.net/wiki`
-   - Name: `CONFLUENCE_EMAIL`, Value: `trades@rabitaitrades.com`
+   - Name: `CONFLUENCE_EMAIL`, Value: `your-email@example.com`
    - Name: `CONFLUENCE_API_TOKEN`, Value: `your-token`
    - Name: `CONFLUENCE_SPACE_KEY`, Value: `DOCS`
    - Name: `CONFLUENCE_PARENT_PAGE_ID`, Value: `123456789` (optional)
@@ -939,13 +949,17 @@ features:
 
 | Field | Type | Required | Description |
 |-------|------|----------|-------------|
-| `configVersion` | number | No | Schema version (current: 1) for future migrations |
+| `configVersion` | number | **Yes** | Schema version (must be `1`) |
 | `project.name` | string | Yes | Project name |
 | `project.docs_title_prefix` | string | No | Prefix for documentation titles (default: project name) |
-| `publishers` | array | Yes | Output targets: `notion`, `confluence`, `markdown` |
+| `publishers` | array | Yes | Output targets: `notion`, `confluence`, `github_wiki`, `markdown` (+ plugin publishers) |
+| `plugins` | array | No | Plugin paths or npm package names |
 | `notion.branches` | array | No | Branch whitelist for Notion publishing. Supports globs. |
 | `notion.includeBranchInTitle` | boolean | No | Add `[branch-name]` to titles (default: `true`) |
 | `confluence.branches` | array | No | Branch whitelist for Confluence publishing. Supports globs. |
+| `github_wiki.branches` | array | No | Branch whitelist for GitHub Wiki publishing. Supports globs. |
+| `github_wiki.sidebar` | boolean | No | Generate `_Sidebar.md` navigation (default: `true`) |
+| `github_wiki.footer` | boolean | No | Generate `_Footer.md` (default: `true`) |
 | `discord.enabled` | boolean | No | Enable Discord notifications (default: `true` if webhook set) |
 | `discord.notifyOn` | string | No | Notification policy: `always`, `significant`, `never` (default: `significant`) |
 | `discord.significantThreshold` | number | No | Change % threshold for notifications (default: `10`) |
@@ -956,7 +970,11 @@ features:
 | `scan.ignore` | array | Yes | Glob patterns to exclude |
 | `module_roots` | array | No | Root directories for module detection |
 | `outputs.pages` | array | Yes | Documentation pages to generate |
-| `features` | object | No | Experimental feature flags |
+| `features` | object | No | Feature flags (boolean values) |
+| `ai.enabled` | boolean | No | Enable AI-powered documentation |
+| `ai.mode` | string | No | AI mode: `hybrid`, `full`, or `off` |
+| `ai.temperature` | number | No | Generation temperature (0–2). Not supported by all models (e.g. gpt-5-mini ignores it) |
+| `ai.max_tokens` | number | No | Max completion tokens per request (>0) |
 
 ---
 
@@ -986,6 +1004,13 @@ Optional for Discord notifications:
 |----------|----------|-------------|
 | `DISCORD_WEBHOOK_URL` | No | Discord webhook URL for team notifications |
 
+Required for GitHub Wiki publisher:
+
+| Variable | Required | Description |
+|----------|----------|-------------|
+| `GITHUB_TOKEN` | Yes | Personal access token or Actions `${{ secrets.GITHUB_TOKEN }}` |
+| `GITHUB_REPOSITORY` | No | `owner/repo` (auto-detected from git remote in Actions) |
+
 **Local Development:** Create `.env` file in project root  
 **GitHub Actions:** Add as repository secrets in Settings → Secrets and variables → Actions
 
@@ -1070,7 +1095,7 @@ npm test
 - Integration workflows
 - Doctor command validation
 
-**Coverage:** 163 tests passing across 14 test files
+**Coverage:** 180 tests passing across 15 test files
 
 ### Test Package Installation Locally
 
@@ -1081,7 +1106,7 @@ Simulates the full user installation experience:
 npm pack
 
 # Install globally from tarball
-npm install -g chappibunny-repolens-0.9.0.tgz
+npm install -g chappibunny-repolens-1.2.0.tgz
 
 # Verify
 repolens --version
@@ -1130,11 +1155,15 @@ repolens/
 │   │   ├── publish.js       # Publishing pipeline
 │   │   ├── notion.js        # Notion API integration
 │   │   ├── confluence.js    # Confluence REST API integration
+│   │   ├── github-wiki.js   # GitHub Wiki publisher (git-based)
 │   │   └── markdown.js      # Local Markdown generation
 │   ├── integrations/
 │   │   └── discord.js       # Discord webhook notifications
 │   ├── delivery/
 │   │   └── comment.js       # PR comment delivery
+│   ├── plugins/
+│   │   ├── loader.js        # Plugin resolution and dynamic import
+│   │   └── manager.js       # Plugin registry and lifecycle orchestration
 │   └── utils/
 │       ├── logger.js        # Logging utilities
 │       ├── retry.js         # API retry logic (exponential backoff)
@@ -1146,10 +1175,11 @@ repolens/
 │       ├── telemetry.js     # Opt-in error tracking + performance timers
 │       ├── errors.js        # Enhanced error messages with guidance
 │       └── update-check.js  # Version update notifications
-├── tests/                   # Vitest test suite (163 tests across 14 files)
+├── tests/                   # Vitest test suite (180 tests across 15 files)
 ├── .repolens.yml            # Dogfooding config
 ├── package.json
 ├── CHANGELOG.md
+├── STABILITY.md
 ├── RELEASE.md
 └── ROADMAP.md
 ```
@@ -1163,13 +1193,13 @@ RepoLens uses automated GitHub Actions releases.
 ### Creating a Release
 
 ```bash
-# Patch version (0.9.0 → 0.9.1) - Bug fixes
+# Patch version (1.0.0 → 1.0.1) - Bug fixes
 npm run release:patch
 
-# Minor version (0.9.0 → 0.10.0) - New features
+# Minor version (1.0.0 → 1.1.0) - New features
 npm run release:minor
 
-# Major version (0.9.0 → 1.0.0) - Breaking changes
+# Major version (1.0.0 → 2.0.0) - Breaking changes
 npm run release:major
 
 # Push the tag to trigger workflow
@@ -1189,26 +1219,25 @@ See [RELEASE.md](./RELEASE.md) for detailed workflow.
 
 ## 🤝 Contributing
 
-RepoLens is currently in early access. v1.0 will open for community contributions.
-
 **Ways to help:**
 - **Try it out**: Install and use in your projects
 - **Report issues**: Share bugs, edge cases, or UX friction
 - **Request features**: Tell us what's missing
-- **Share feedback**: What works? What doesn't?
+- **Build plugins**: Extend RepoLens with custom renderers and publishers
+- **Share feedback**: `repolens feedback`
 
 ---
 
-## 🗺️ Roadmap to v1.0
+## 🗺️ Roadmap
 
-**Current Status:** v0.9.0 — Plugin System
+**Current Status:** v1.2.0 — Stable Release
 
-### Completed ✅
+### v1.0 — Complete ✅
 
 - [x] CLI commands: `init`, `doctor`, `publish`, `migrate`, `watch`, `feedback`, `version`, `help`
-- [x] Config schema v1 with validation
+- [x] Config schema v1 with validation (frozen)
 - [x] Auto-discovery of `.repolens.yml`
-- [x] Publishers: Notion + Confluence + Markdown
+- [x] Publishers: Notion + Confluence + GitHub Wiki + Markdown
 - [x] Branch-aware publishing with filtering
 - [x] Smart tech stack detection from package.json
 - [x] Unicode dependency diagrams (no external deps)
@@ -1217,7 +1246,7 @@ RepoLens is currently in early access. v1.0 will open for community contribution
 - [x] GitHub Actions automation (publish + release)
 - [x] PR architecture diff comments
 - [x] Performance guardrails (10k warning, 50k limit)
-- [x] Comprehensive test suite (163 tests across 14 files)
+- [x] Comprehensive test suite (180 tests across 15 files)
 - [x] Security hardening (secret detection, injection prevention, fuzzing)
 - [x] Discord notifications with rich embeds
 - [x] Documentation coverage & health scoring
@@ -1235,11 +1264,14 @@ RepoLens is currently in early access. v1.0 will open for community contribution
 - [x] Dependency graph with circular dependency detection
 - [x] Architecture drift detection (baseline comparison)
 - [x] Plugin system for custom renderers and publishers
+- [x] Stability audit: CLI, config schema, and plugin interface frozen
+- [x] [STABILITY.md](STABILITY.md): Public API contract with semver guarantees
 
-### Planned for v1.0 🎯
+### Future
 
-- [ ] Stability audit: freeze CLI flags and config schema
-- [ ] Additional publishers (GitHub Wiki, Obsidian)
+- [ ] Additional publishers (Obsidian)
+- [ ] VS Code extension
+- [ ] GitHub App
 
 See [ROADMAP.md](./ROADMAP.md) for detailed planning.
 
@@ -1253,6 +1285,8 @@ MIT
 
 ## 💬 Support & Contact
 
+- **Troubleshooting**: [TROUBLESHOOTING.md](TROUBLESHOOTING.md) — installation, config, publishing, AI, and CI/CD issues
+- **Diagnostics**: Run `npx @chappibunny/repolens doctor` to validate your setup
 - **Issues**: [GitHub Issues](https://github.com/CHAPIBUNNY/repolens/issues)
 - **Discussions**: [GitHub Discussions](https://github.com/CHAPIBUNNY/repolens/discussions)
 - **Email**: Contact repository maintainers

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@chappibunny/repolens",
-  "version": "0.9.0",
+  "version": "1.2.0",
   "description": "AI-assisted documentation intelligence system for technical and non-technical audiences",
   "license": "MIT",
   "type": "module",

package/src/ai/generate-sections.js

@@ -25,7 +25,6 @@ export async function generateExecutiveSummary(context) {
   const result = await generateText({
     system: SYSTEM_PROMPT,
     user: createExecutiveSummaryPrompt(context),
-    temperature: 0.2,
     maxTokens: 1500
   });
   
@@ -47,7 +46,6 @@ export async function generateSystemOverview(context) {
   const result = await generateText({
     system: SYSTEM_PROMPT,
     user: createSystemOverviewPrompt(context),
-    temperature: 0.2,
     maxTokens: 1200
   });
   
@@ -68,7 +66,6 @@ export async function generateBusinessDomains(context) {
   const result = await generateText({
     system: SYSTEM_PROMPT,
     user: createBusinessDomainsPrompt(context),
-    temperature: 0.2,
     maxTokens: 2000
   });
   
@@ -89,7 +86,6 @@ export async function generateArchitectureOverview(context) {
   const result = await generateText({
     system: SYSTEM_PROMPT,
     user: createArchitectureOverviewPrompt(context),
-    temperature: 0.2,
     maxTokens: 1800
   });
   
@@ -110,7 +106,6 @@ export async function generateDataFlows(flows, context) {
   const result = await generateText({
     system: SYSTEM_PROMPT,
     user: createDataFlowsPrompt(flows, context),
-    temperature: 0.2,
     maxTokens: 1800
   });
   
@@ -131,7 +126,6 @@ export async function generateDeveloperOnboarding(context) {
   const result = await generateText({
     system: SYSTEM_PROMPT,
     user: createDeveloperOnboardingPrompt(context),
-    temperature: 0.2,
     maxTokens: 2200
   });
   

package/src/ai/provider.js

@@ -4,12 +4,12 @@ import { warn, info } from "../utils/logger.js";
 import { executeAIRequest } from "../utils/rate-limit.js";
 
 const DEFAULT_TIMEOUT_MS = 60000;
-const DEFAULT_TEMPERATURE = 0.2;
 const DEFAULT_MAX_TOKENS = 2500;
 
-export async function generateText({ system, user, temperature, maxTokens }) {
-  // Check if AI is enabled
-  const enabled = process.env.REPOLENS_AI_ENABLED === "true";
+export async function generateText({ system, user, temperature, maxTokens, config }) {
+  // Check if AI is enabled (env var takes precedence, then config)
+  const aiConfig = config?.ai || {};
+  const enabled = process.env.REPOLENS_AI_ENABLED === "true" || aiConfig.enabled === true;
   
   if (!enabled) {
     return {
@@ -19,13 +19,17 @@ export async function generateText({ system, user, temperature, maxTokens }) {
     };
   }
   
-  // Get provider configuration from environment
+  // Get provider configuration (env vars take precedence, then config, then defaults)
   const provider = process.env.REPOLENS_AI_PROVIDER || "openai_compatible";
   const baseUrl = process.env.REPOLENS_AI_BASE_URL;
   const apiKey = process.env.REPOLENS_AI_API_KEY;
-  const model = process.env.REPOLENS_AI_MODEL || "gpt-4-turbo-preview";
+  const model = process.env.REPOLENS_AI_MODEL || "gpt-5-mini";
   const timeoutMs = parseInt(process.env.REPOLENS_AI_TIMEOUT_MS || DEFAULT_TIMEOUT_MS);
   
+  // Use config values as fallback for maxTokens; temperature only when explicitly set
+  const resolvedTemp = temperature ?? aiConfig.temperature ?? undefined;
+  const resolvedMaxTokens = maxTokens ?? aiConfig.max_tokens ?? DEFAULT_MAX_TOKENS;
+  
   // Validate configuration
   if (!apiKey) {
     warn("REPOLENS_AI_API_KEY not set. AI features disabled.");
@@ -47,8 +51,8 @@ export async function generateText({ system, user, temperature, maxTokens }) {
       model,
       system,
       user,
-      temperature: temperature ?? DEFAULT_TEMPERATURE,
-      maxTokens: maxTokens ?? DEFAULT_MAX_TOKENS,
+      temperature: resolvedTemp,
+      maxTokens: resolvedMaxTokens,
       timeoutMs
     });
     
@@ -76,21 +80,28 @@ async function callOpenAICompatibleAPI({ baseUrl, apiKey, model, system, user, t
     const timeoutId = setTimeout(() => controller.abort(), timeoutMs);
     
     try {
+      // Build request body — omit temperature if not supported by model
+      const body = {
+        model,
+        messages: [
+          { role: "system", content: system },
+          { role: "user", content: user }
+        ],
+        max_completion_tokens: maxTokens
+      };
+      // Only send temperature when explicitly configured — some models
+      // (e.g. gpt-5-mini) reject any non-default value
+      if (temperature != null) {
+        body.temperature = temperature;
+      }
+
       const response = await fetch(url, {
         method: "POST",
         headers: {
           "Content-Type": "application/json",
           "Authorization": `Bearer ${apiKey}`
         },
-        body: JSON.stringify({
-          model,
-          messages: [
-            { role: "system", content: system },
-            { role: "user", content: user }
-          ],
-          temperature,
-          max_tokens: maxTokens
-        }),
+        body: JSON.stringify(body),
         signal: controller.signal
       });
       
@@ -129,9 +140,9 @@ export function getAIConfig() {
   return {
     enabled: isAIEnabled(),
     provider: process.env.REPOLENS_AI_PROVIDER || "openai_compatible",
-    model: process.env.REPOLENS_AI_MODEL || "gpt-4-turbo-preview",
+    model: process.env.REPOLENS_AI_MODEL || "gpt-5-mini",
     hasApiKey: !!process.env.REPOLENS_AI_API_KEY,
-    temperature: parseFloat(process.env.REPOLENS_AI_TEMPERATURE || DEFAULT_TEMPERATURE),
+    temperature: process.env.REPOLENS_AI_TEMPERATURE ? parseFloat(process.env.REPOLENS_AI_TEMPERATURE) : undefined,
     maxTokens: parseInt(process.env.REPOLENS_AI_MAX_TOKENS || DEFAULT_MAX_TOKENS)
   };
 }