forgecodecli 0.1.0__tar.gz → 0.2.0__tar.gz

forgecodecli-0.2.0/PKG-INFO

@@ -0,0 +1,198 @@
+Metadata-Version: 2.4
+Name: forgecodecli
+Version: 0.2.0
+Summary: A minimal agentic CLI for forging code
+Author: Sudhanshu
+License: MIT
+Requires-Python: >=3.9
+Description-Content-Type: text/markdown
+Requires-Dist: typer[all]>=0.9.0
+Requires-Dist: openai>=1.0.0
+Requires-Dist: keyring>=24.0.0
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0.0; extra == "dev"
+Requires-Dist: black>=23.0.0; extra == "dev"
+Requires-Dist: ruff>=0.1.0; extra == "dev"
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=0.25.0; extra == "anthropic"
+
+# ForgeCodeCLI
+
+An agentic, file-aware command-line tool that lets you manage and modify your codebase using natural language — powered by LLMs.
+
+It acts as a safe, deterministic AI agent that can read files, create/delete directories, and write code only through explicit tools, not raw hallucination.
+
+## Features
+
+- ✅ **Agentic workflow** - LLM decides actions, CLI executes them safely
+- ✅ **File operations** - Read, list, create, write, delete, move files & directories
+- ✅ **Undo support** - Reverse the last file operation with `undo`
+- ✅ **Multi-provider LLMs** - Gemini, OpenAI (GPT), Anthropic (Claude), Groq
+- ✅ **Model selection** - Choose specific models for each provider
+- ✅ **Secure storage** - API keys stored in system keyring (no env vars)
+- ✅ **Deterministic** - Rule-based execution with validation
+- ✅ **Interactive CLI** - Real-time agent feedback
+
+## Installation
+
+Requires Python 3.9+
+
+```bash
+pip install forgecodecli
+```
+
+**Optional:** For Anthropic (Claude) support, install with the anthropic extra:
+
+```bash
+pip install forgecodecli[anthropic]
+```
+
+Or install later when prompted during setup.
+
+## Quick Start
+
+### Initialize (one-time setup)
+
+```bash
+forgecodecli init
+```
+
+You will be prompted to:
+
+1. **Select LLM Provider**
+
+   ```
+   1) Google Gemini
+   2) OpenAI
+   3) Anthropic (Claude)
+   4) Groq
+   ```
+
+2. **Select Model** (varies by provider)
+   - Gemini: `gemini-2.5-flash`, `gemini-2.0-flash`, `gemini-1.5-pro`
+   - OpenAI: `gpt-4o`, `gpt-4-turbo`, `gpt-3.5-turbo`
+   - Claude: `claude-3-5-sonnet`, `claude-3-opus`, `claude-3-haiku`
+   - Groq: `llama-3.3-70b`, `mixtral-8x7b`, `gemma2-9b-it`
+
+3. **Enter API Key** (stored securely in system keyring)
+
+### Start the agent
+
+```bash
+forgecodecli
+```
+
+You are now in interactive agent mode. Example commands:
+
+```
+create a folder src/app with main.py that prints "Hello World"
+read the config.py file
+list all files in src
+delete old_backup folder
+move test.py to tests/test.py
+undo
+```
+
+Type `help` for built-in commands, or press `Ctrl + C` to exit.
+
+## Reset Configuration
+
+To remove all configuration and API keys:
+
+```bash
+forgecodecli reset
+```
+
+## Security
+
+- API keys are stored using the system keyring
+- No API keys are written to config files or environment variables
+- Config files contain only non-sensitive metadata
+
+## How It Works
+
+1. You enter a natural language command
+2. The LLM decides the next valid action
+3. ForgeCodeCLI executes the action with validation
+4. The agent receives feedback and responds
+5. Process repeats until agent provides an answer
+
+**Safety mechanisms:**
+
+- Action limit of 2 per request (prevents infinite loops)
+- Conversation context maintained for agent awareness
+- All operations logged and reversible with `undo`
+- Strict tool validation
+
+## Supported Actions
+
+The agent can execute these operations:
+
+**File Operations:**
+| Action | Description |
+|--------|-------------|
+| `read_file` | Read and display file contents |
+| `list_files` | List files in a directory |
+| `create_dir` | Create new directories |
+| `write_file` | Create and write files |
+| `delete_file` | Delete files permanently |
+| `delete_dir` | Delete directories |
+| `move_file` | Move or rename files |
+| `move_dir` | Move or rename directories |
+| `undo` | Reverse the last operation |
+
+**Git Operations:**
+| Action | Description |
+|--------|-------------|
+| `git_init` | Initialize git repository |
+| `git_add` | Stage files for commit |
+| `git_commit` | Commit staged changes |
+| `git_push` | Push to remote repository |
+| `git_set_origin` | Set remote repository URL |
+| `git_status` | Show repository status |
+| `git_log` | View commit history |
+| `git_branch` | Manage branches |
+| `git_pull` | Pull from remote |
+| `git_clone` | Clone a repository |
+
+All actions are executed safely with validation and error handling.
+
+## Roadmap
+
+### ✅ v1 (Released)
+
+- Basic file operations (read, list, create, write)
+- Gemini support
+- Interactive CLI
+
+### ✅ v2 (Released)
+
+- **Undo functionality** - Stack-based operation reversal
+- **Delete & move operations** - Full file/directory manipulation
+- **Multi-provider support** - Gemini, OpenAI, Anthropic, Groq
+- **Model selection** - Choose specific models per provider
+- **Auto-install SDKs** - Anthropic SDK installs on demand
+- **Fixed agent loop** - Proper conversation flow with max actions
+
+### ✅ v3 (Current)
+
+- **Git operations** - init, add, commit, push, pull, status, log, branch, clone, set_origin
+- **Git workflow support** - Full version control integration
+- **Repository management** - Clone and manage git repos
+
+### 🚀 v4 (Planned)
+
+- Copy files/directories
+- Full undo/redo history (not just last operation)
+- File search capabilities
+- Code generation templates
+- Batch operations
+- Backup/snapshot functionality
+
+## License
+
+MIT License
+
+## Author
+
+Built by Sudhanshu

forgecodecli-0.2.0/README.md

@@ -0,0 +1,180 @@
+# ForgeCodeCLI
+
+An agentic, file-aware command-line tool that lets you manage and modify your codebase using natural language — powered by LLMs.
+
+It acts as a safe, deterministic AI agent that can read files, create/delete directories, and write code only through explicit tools, not raw hallucination.
+
+## Features
+
+- ✅ **Agentic workflow** - LLM decides actions, CLI executes them safely
+- ✅ **File operations** - Read, list, create, write, delete, move files & directories
+- ✅ **Undo support** - Reverse the last file operation with `undo`
+- ✅ **Multi-provider LLMs** - Gemini, OpenAI (GPT), Anthropic (Claude), Groq
+- ✅ **Model selection** - Choose specific models for each provider
+- ✅ **Secure storage** - API keys stored in system keyring (no env vars)
+- ✅ **Deterministic** - Rule-based execution with validation
+- ✅ **Interactive CLI** - Real-time agent feedback
+
+## Installation
+
+Requires Python 3.9+
+
+```bash
+pip install forgecodecli
+```
+
+**Optional:** For Anthropic (Claude) support, install with the anthropic extra:
+
+```bash
+pip install forgecodecli[anthropic]
+```
+
+Or install later when prompted during setup.
+
+## Quick Start
+
+### Initialize (one-time setup)
+
+```bash
+forgecodecli init
+```
+
+You will be prompted to:
+
+1. **Select LLM Provider**
+
+   ```
+   1) Google Gemini
+   2) OpenAI
+   3) Anthropic (Claude)
+   4) Groq
+   ```
+
+2. **Select Model** (varies by provider)
+   - Gemini: `gemini-2.5-flash`, `gemini-2.0-flash`, `gemini-1.5-pro`
+   - OpenAI: `gpt-4o`, `gpt-4-turbo`, `gpt-3.5-turbo`
+   - Claude: `claude-3-5-sonnet`, `claude-3-opus`, `claude-3-haiku`
+   - Groq: `llama-3.3-70b`, `mixtral-8x7b`, `gemma2-9b-it`
+
+3. **Enter API Key** (stored securely in system keyring)
+
+### Start the agent
+
+```bash
+forgecodecli
+```
+
+You are now in interactive agent mode. Example commands:
+
+```
+create a folder src/app with main.py that prints "Hello World"
+read the config.py file
+list all files in src
+delete old_backup folder
+move test.py to tests/test.py
+undo
+```
+
+Type `help` for built-in commands, or press `Ctrl + C` to exit.
+
+## Reset Configuration
+
+To remove all configuration and API keys:
+
+```bash
+forgecodecli reset
+```
+
+## Security
+
+- API keys are stored using the system keyring
+- No API keys are written to config files or environment variables
+- Config files contain only non-sensitive metadata
+
+## How It Works
+
+1. You enter a natural language command
+2. The LLM decides the next valid action
+3. ForgeCodeCLI executes the action with validation
+4. The agent receives feedback and responds
+5. Process repeats until agent provides an answer
+
+**Safety mechanisms:**
+
+- Action limit of 2 per request (prevents infinite loops)
+- Conversation context maintained for agent awareness
+- All operations logged and reversible with `undo`
+- Strict tool validation
+
+## Supported Actions
+
+The agent can execute these operations:
+
+**File Operations:**
+| Action | Description |
+|--------|-------------|
+| `read_file` | Read and display file contents |
+| `list_files` | List files in a directory |
+| `create_dir` | Create new directories |
+| `write_file` | Create and write files |
+| `delete_file` | Delete files permanently |
+| `delete_dir` | Delete directories |
+| `move_file` | Move or rename files |
+| `move_dir` | Move or rename directories |
+| `undo` | Reverse the last operation |
+
+**Git Operations:**
+| Action | Description |
+|--------|-------------|
+| `git_init` | Initialize git repository |
+| `git_add` | Stage files for commit |
+| `git_commit` | Commit staged changes |
+| `git_push` | Push to remote repository |
+| `git_set_origin` | Set remote repository URL |
+| `git_status` | Show repository status |
+| `git_log` | View commit history |
+| `git_branch` | Manage branches |
+| `git_pull` | Pull from remote |
+| `git_clone` | Clone a repository |
+
+All actions are executed safely with validation and error handling.
+
+## Roadmap
+
+### ✅ v1 (Released)
+
+- Basic file operations (read, list, create, write)
+- Gemini support
+- Interactive CLI
+
+### ✅ v2 (Released)
+
+- **Undo functionality** - Stack-based operation reversal
+- **Delete & move operations** - Full file/directory manipulation
+- **Multi-provider support** - Gemini, OpenAI, Anthropic, Groq
+- **Model selection** - Choose specific models per provider
+- **Auto-install SDKs** - Anthropic SDK installs on demand
+- **Fixed agent loop** - Proper conversation flow with max actions
+
+### ✅ v3 (Current)
+
+- **Git operations** - init, add, commit, push, pull, status, log, branch, clone, set_origin
+- **Git workflow support** - Full version control integration
+- **Repository management** - Clone and manage git repos
+
+### 🚀 v4 (Planned)
+
+- Copy files/directories
+- Full undo/redo history (not just last operation)
+- File search capabilities
+- Code generation templates
+- Batch operations
+- Backup/snapshot functionality
+
+## License
+
+MIT License
+
+## Author
+
+Built by Sudhanshu

forgecodecli-0.2.0/forgecodecli/agent.py

@@ -0,0 +1,215 @@
+# from dotenv import load_dotenv
+from openai import OpenAI
+from openai import RateLimitError
+import json
+from json import JSONDecoder
+from forgecodecli.config import load_config
+from forgecodecli.secrets import load_api_key
+from forgecodecli.tools import undo
+from forgecodecli.git_tools import (
+    git_init, git_add, git_commit, git_push, git_set_origin,
+    git_status, git_log, git_branch, git_pull, git_clone
+)
+
+# Provider base URLs (OpenAI-compatible)
+PROVIDER_BASE_URLS = {
+    "gemini":    "https://generativelanguage.googleapis.com/v1beta/openai/",
+    "openai":    None,   # default OpenAI
+    "groq":      "https://api.groq.com/openai/v1",
+}
+
+def get_client():
+    config = load_config()
+    api_key = load_api_key()
+
+    if not config:
+        raise RuntimeError(
+            "ForgeCodeCLI is not set up. Run `forgecodecli init`."
+        )
+
+    if not api_key:
+        raise RuntimeError(
+            "API key not found. Run `forgecodecli init` again."
+        )
+
+    provider = config.get("provider", "gemini")
+
+    if provider == "anthropic":
+        try:
+            import anthropic as _anthropic
+            return _anthropic.Anthropic(api_key=api_key)
+        except ImportError:
+            raise RuntimeError(
+                "Anthropic package not installed. Run: pip install anthropic"
+            )
+
+    base_url = PROVIDER_BASE_URLS.get(provider)
+    if base_url:
+        return OpenAI(api_key=api_key, base_url=base_url)
+    return OpenAI(api_key=api_key)
+
+SYSTEM_PROMPT = """
+You are a file and git management agent that executes file and version control operations. You MUST use the actions below to fulfill user requests. You respond ONLY with valid JSON.
+
+═══════════════════════════════════════════════════════════════
+ABSOLUTE RULES - VIOLATIONS ARE FORBIDDEN
+═══════════════════════════════════════════════════════════════
+
+- You CAN delete files. You CAN delete directories. You CAN move files. You CAN move directories.
+- NEVER say "I cannot", "I'm unable to", "I don't have the ability to", or any variation.
+- NEVER refuse a delete, move, or any file operation. You MUST execute it.
+- When a user asks to delete a file, you MUST respond with the delete_file action.
+- When a user asks to delete a folder, you MUST respond with the delete_dir action.
+
+═══════════════════════════════════════════════════════════════
+ALL AVAILABLE ACTIONS
+═══════════════════════════════════════════════════════════════
+
+FILE OPERATIONS:
+1. "read_file"   → Read file contents.        Args: {"path": "..."}
+2. "list_files"  → List directory contents.    Args: {"path": "..."}
+3. "write_file"  → Create/write a file.        Args: {"path": "...", "content": "..."}
+4. "create_dir"  → Create a directory.         Args: {"path": "..."}
+5. "delete_file" → DELETE a file permanently.   Args: {"path": "..."}
+6. "delete_dir"  → DELETE an empty directory.   Args: {"path": "..."}
+7. "move_file"   → Move or rename a file.      Args: {"src": "...", "dst": "..."}
+8. "move_dir"    → Move or rename directory.    Args: {"src": "...", "dst": "..."}
+9. "undo"        → Undo the last operation.    Args: {}
+
+GIT OPERATIONS:
+10. "git_init"       → Initialize git repo.           Args: {}
+11. "git_add"        → Stage files.                    Args: {"path": "." or specific file}
+12. "git_commit"     → Commit changes.                 Args: {"message": "commit message"}
+13. "git_push"       → Push to remote.                 Args: {"branch": "main" (default)}
+14. "git_set_origin" → Set remote URL.                 Args: {"url": "https://..."}
+15. "git_status"     → Show status.                    Args: {}
+16. "git_log"        → Show commit history.            Args: {"lines": 5 (default)}
+17. "git_branch"     → Manage branches.                Args: {"name": "branch_name" (optional)}
+18. "git_pull"       → Pull from remote.               Args: {}
+19. "git_clone"      → Clone repository.               Args: {"url": "...", "path": "." (default)}
+20. "answer"         → Respond to user with text.     Args: {"text": "..."}
+
+═══════════════════════════════════════════════════════════════
+CRITICAL RULES
+═══════════════════════════════════════════════════════════════
+
+1. MAXIMUM 2 actions per request (not counting "answer")
+2. After completing all required actions, ALWAYS respond with "answer"
+3. Do NOT repeat the same action twice
+4. Do NOT take multiple write_file or create_dir actions in one request
+5. Choose the MOST DIRECT action for the task
+
+═══════════════════════════════════════════════════════════════
+RESPONSE FORMAT - ONLY VALID JSON
+═══════════════════════════════════════════════════════════════
+
+You MUST respond ONLY with valid JSON, nothing else. No text before or after.
+
+{"action": "action_name", "args": {"key": "value"}}
+
+═══════════════════════════════════════════════════════════════
+EXAMPLES
+═══════════════════════════════════════════════════════════════
+
+User: "delete love.txt from megha folder"
+→ {"action": "delete_file", "args": {"path": "megha/love.txt"}}
+Then: {"action": "answer", "args": {"text": "File megha/love.txt deleted."}}
+
+User: "initialize git and commit"
+→ {"action": "git_init", "args": {}}
+Then: {"action": "git_add", "args": {"path": "."}}
+Then: {"action": "git_commit", "args": {"message": "Initial commit"}}
+Then: {"action": "answer", "args": {"text": "Repository initialized and committed."}}
+
+User: "push to origin"
+→ {"action": "git_push", "args": {"branch": "main"}}
+Then: {"action": "answer", "args": {"text": "Pushed to origin/main."}}
+
+User: "create hello.py with print('hi')"
+→ {"action": "write_file", "args": {"path": "hello.py", "content": "print('hi')"}}
+Then: {"action": "answer", "args": {"text": "File created."}}
+
+REMEMBER: You have delete_file, delete_dir, undo, and git tools. USE THEM when asked.
+"""
+
+def think(messages: list[dict]) -> dict:
+    config = load_config()
+    model = config.get("model", "gemini-2.5-flash")
+    provider = config.get("provider", "gemini")
+    try:
+      client = get_client()
+
+      # Anthropic uses its own SDK format
+      if provider == "anthropic":
+          response = client.messages.create(
+              model=model,
+              max_tokens=1024,
+              system=SYSTEM_PROMPT,
+              messages=messages,
+          )
+          content = response.content[0].text
+      else:
+          response = client.chat.completions.create(
+              model=model,
+              messages=[{"role": "system", "content": SYSTEM_PROMPT}] + messages,
+              temperature=0
+          )
+          content = response.choices[0].message.content
+
+    except RateLimitError as e:
+      return {
+        "action": "answer",
+        "args": {
+            "text": "⚠️ Rate limit hit. Please wait a few seconds and try again."
+        }
+    }
+    except Exception as e:
+      return {
+        "action": "answer",
+        "args": {
+            "text": f"❌ LLM error: {str(e)}"
+        }
+    }
+    
+    # Handle empty response
+    if content is None or not content.strip():
+        return {
+            "action": "answer",
+            "args": {
+                "text": "Task completed successfully!"
+            }
+        }
+    
+    cleaned = content.strip()
+
+    # Robust JSON extraction
+    decoder = JSONDecoder()
+    
+    # Try to find JSON object in the content
+    idx = cleaned.find("{")
+    if idx == -1:
+        return {
+        "action": "answer",
+        "args": {
+            "text": cleaned
+        }
+    }
+    
+    # Extract from first { onwards
+    cleaned = cleaned[idx:]
+    
+    # Try to decode JSON, handling partial/malformed content
+    try:
+        obj, _ = decoder.raw_decode(cleaned)
+        return obj
+    except json.JSONDecodeError:
+        # If it fails, try to find the end of a valid JSON object
+        # by trying progressively shorter strings from the end
+        for end_pos in range(len(cleaned), idx, -1):
+            try:
+                obj, _ = decoder.raw_decode(cleaned[:end_pos])
+                return obj
+            except json.JSONDecodeError:
+                continue
+        
+        raise ValueError(f"Could not parse JSON from LLM output: {cleaned[:100]}...")