Stacker cli, minimal requirements for deployment.

.github/workflows/rust.yml

@@ -67,11 +67,16 @@ jobs:
 
       - name: Build console (release with features)
         run: cargo build --release --target ${{ matrix.target }} --bin console --features explain --verbose
+
+      - name: Build stacker-cli (release)
+        run: cargo build --release --target ${{ matrix.target }} --bin stacker-cli --verbose
+
       - name: Prepare binaries
         run: |
           mkdir -p artifacts
           cp target/${{ matrix.target }}/release/server artifacts/server
           cp target/${{ matrix.target }}/release/console artifacts/console
+          cp target/${{ matrix.target }}/release/stacker-cli artifacts/stacker-cli
           tar -czf ${{ matrix.artifact_name }}.tar.gz -C artifacts .
       - name: Upload binaries
         uses: actions/upload-artifact@v4

CHANGELOG.md

@@ -2,6 +2,97 @@
 
 All notable changes to this project will be documented in this file.
 
+## [0.2.3] — 2026-02-23
+
+### Changed — `stacker init` now generates `.stacker/` directory
+
+- `stacker init` now creates `.stacker/Dockerfile` and `.stacker/docker-compose.yml` alongside `stacker.yml`, so the project is ready to deploy immediately without running `deploy --dry-run` first
+- Dockerfile generation is skipped when `app.image` or `app.dockerfile` is set in the config
+- Compose generation is skipped when `deploy.compose_file` is set
+
+### Changed — `stacker deploy` reuses existing `.stacker/` artifacts
+
+- `deploy` no longer errors when `.stacker/Dockerfile` or `.stacker/docker-compose.yml` already exist (e.g. from `stacker init`)
+- Existing artifacts are reused; pass `--force-rebuild` to regenerate them
+
+### Added — `--ai-provider`, `--ai-model`, `--ai-api-key` flags on `stacker-cli init`
+
+- The `stacker-cli` binary (`console/main.rs`) now supports all AI-related flags that the standalone `stacker` binary already had:
+  - `--ai-provider <PROVIDER>` — openai, anthropic, ollama, custom
+  - `--ai-model <MODEL>` — e.g. `qwen2.5-coder`, `deepseek-r1`, `gpt-4o`
+  - `--ai-api-key <KEY>` — API key for cloud AI providers
+
+### Added — AI troubleshooting suggestions on deploy failures
+
+- On `stacker deploy` failures (`DeployFailed`), CLI now attempts AI-assisted troubleshooting automatically
+- It sends the deploy error plus generated `.stacker/Dockerfile` and `.stacker/docker-compose.yml` snippets to the configured AI provider
+- If AI is unavailable or not configured, CLI prints deterministic fallback hints for common issues (for example `npm ci` failures, obsolete compose `version`, missing files, permissions, and network timeouts)
+
+### Fixed
+
+- `stacker-cli init --with-ai --ai-model qwen2.5-coder` no longer fails with an unrecognised flag error
+- `stacker deploy` after `stacker init` no longer fails with `DockerfileExists` error
+
+## 2026-02-23
+
+### Added - Configurable AI Request Timeout
+
+- New `timeout` field in `ai` config section of `stacker.yml` (default: 300 seconds)
+- `STACKER_AI_TIMEOUT` environment variable overrides the config value
+- Timeout applies to all AI providers (OpenAI, Anthropic, Ollama, Custom)
+- Useful for large models on slower hardware: `STACKER_AI_TIMEOUT=900 stacker init --with-ai`
+- Example stacker.yml:
+  ```yaml
+  ai:
+    enabled: true
+    provider: ollama
+    model: deepseek-r1
+    timeout: 600  # 10 minutes
+  ```
+- 9 new tests for timeout resolution
+
+### Added - Stacker CLI: AI-Powered Project Initialization
+
+#### AI Scanner Module (`src/cli/ai_scanner.rs`)
+- New `scan_project()` function performs deep project scanning, reading key config files (`package.json`, `requirements.txt`, `Cargo.toml`, `Dockerfile`, `docker-compose.yml`, `.env`, etc.) to build rich context for AI generation
+- `build_generation_prompt()` constructs detailed prompts including detected app type, file contents, existing infrastructure, and env var keys (values redacted for security)
+- `generate_config_with_ai()` sends project context to the configured AI provider and returns a tailored `stacker.yml`
+- `strip_code_fences()` strips markdown code fences from AI responses
+- System prompt encodes the full `stacker.yml` schema so the AI generates valid, deployable configs
+- 16 unit tests
+
+#### AI-Powered `stacker init --with-ai` (`src/console/commands/cli/init.rs`)
+- `stacker init --with-ai` now scans the project and calls the AI to generate a tailored `stacker.yml` with appropriate services, proxy, monitoring, and hooks
+- New CLI flags on `stacker init`:
+  - `--ai-provider <PROVIDER>` — AI provider: `openai`, `anthropic`, `ollama`, `custom` (default: `ollama`)
+  - `--ai-model <MODEL>` — Model name (e.g. `gpt-4o`, `claude-sonnet-4-20250514`, `llama3`)
+  - `--ai-api-key <KEY>` — API key (or use environment variables)
+- `resolve_ai_config()` resolves AI configuration with priority: CLI flag → environment variable → defaults
+- Environment variable support: `STACKER_AI_PROVIDER`, `STACKER_AI_MODEL`, `STACKER_AI_API_KEY`, `OPENAI_API_KEY`, `ANTHROPIC_API_KEY`
+- Graceful fallback: if AI generation fails (provider unreachable, invalid YAML), automatically falls back to template-based generation
+- AI-generated configs include a review header noting the provider and model used
+- If AI output fails validation, raw draft is saved to `stacker.yml.ai-draft` for manual review
+- 8 new unit tests (18 total in init.rs), 3 new integration tests (11 total in cli_init.rs)
+
+#### Usage Examples
+```bash
+# AI-powered init with local Ollama
+stacker init --with-ai
+
+# AI-powered init with OpenAI
+stacker init --with-ai --ai-provider openai --ai-api-key sk-...
+
+# AI-powered init with Anthropic (key from env)
+export ANTHROPIC_API_KEY=sk-ant-...
+stacker init --with-ai --ai-provider anthropic
+
+# Falls back to template if AI fails
+stacker init --with-ai  # no Ollama running → template fallback
+```
+
+### Test Results
+- **467 tests** (426 unit + 41 integration), 0 failures
+
 ## 2026-02-18
 
 ### Fixed