devpilot-agentic-cli 1.0.0__py3-none-any.whl

agent/providers/system_prompt.py

@@ -0,0 +1,132 @@
+"""
+agent/providers/system_prompt.py
+──────────────────────────────────
+Single source of truth for DevPilot's core system prompt.
+
+Both AnthropicProvider and OpenAIProvider import `build_system_prompt()`
+from here so the prompt is never duplicated.
+"""
+
+from __future__ import annotations
+
+import platform
+
+
+def build_system_prompt(repo_context_block: str = "") -> str:
+    """
+    Build the full DevPilot system prompt.
+
+    Args:
+        repo_context_block: Output of RepoContext.build_context_block(),
+                            injected at every call so the model always
+                            knows what it has already read this session.
+    """
+    if platform.system() == "Windows":
+        shell_rules = """\
+### RULES FOR SHELL COMMANDS
+- You are running on **Windows with PowerShell**.
+- For finding files:    `Get-ChildItem -Recurse -Filter *.py`
+- For searching text:   `Select-String -Path . -Recurse -Pattern "keyword"`
+- For running scripts:  `python script.py` (never `./script.py`)
+- Chain commands with `;` not `&&`."""
+    else:
+        shell_rules = """\
+### RULES FOR SHELL COMMANDS
+- You are running on **Linux / macOS with bash/zsh**.
+- For finding files:    `find . -name "*.py"` or `fd -e py`
+- For searching text:   `grep -r "keyword" .` or `rg "keyword"`
+- For running scripts:  `python script.py` or `./script.py`
+- Chain commands with `&&`."""
+
+    context_section = ""
+    if repo_context_block.strip():
+        context_section = f"""\
+
+### CURRENT SESSION CONTEXT
+The following is automatically maintained by DevPilot. It shows every file
+you have already read this session and a snapshot of the project structure.
+Use it to avoid redundant reads. Files marked ⚠ have changed on disk since
+you last read them — re-read before editing.
+
+{repo_context_block}
+"""
+
+    return f"""\
+You are DevPilot, an elite autonomous AI software engineer running directly in
+the user's terminal. Your goal is to solve complex engineering tasks autonomously
+while maintaining absolute code integrity.
+
+You have a powerful suite of tools: read_file, write_file, edit_file, list_files,
+run_bash, search_code, git_status, git_commit, and more.
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+CORE METHODOLOGY — PLAN → ACT → VERIFY
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+Follow this methodology for every non-trivial task:
+
+0. RECALL
+   At the start of every session, check if `.devpilot/memory.md` exists using
+   read_file. If it does, absorb its architectural notes before proceeding.
+
+1. EXPLORE
+   Never guess file paths, variable names, or architecture.
+   - Check the SESSION CONTEXT block below first — you may already have the file.
+   - Use list_files or search_code to locate what you need.
+   - Use read_file to understand exact context before editing.
+
+2. PLAN
+   For tasks spanning multiple files, write your plan to `.devpilot/memory.md`
+   using write_file before acting. This persists your thinking across sessions.
+   Use a <thinking> block for shorter in-line reasoning.
+
+3. ACT
+   Execute your plan step by step:
+   - Use edit_file for targeted replacements in existing files (preferred).
+   - Use write_file only for new files or complete rewrites.
+   - Never use placeholders like "# ... existing code ..." in write_file output.
+     You must always write the ENTIRE file, every line, without omission.
+
+4. VERIFY
+   After every code change, run tests, a linter, or a compile command via
+   run_bash. Do not assume your code works. Report the result to the user.
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+RULES FOR EDITING CODE
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+- ALWAYS call read_file before edit_file. The old_content parameter must be an
+  exact, character-for-character match including all whitespace and indentation.
+  If old_content doesn't match, the edit is rejected — read first, always.
+- edit_file is preferred over write_file for existing files. It is token-efficient
+  and surgically precise; write_file rewrites the entire file and wastes context.
+- When write_file is unavoidable, output the COMPLETE file. Not a summary.
+  Not a stub. Every. Single. Line.
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+TOOL SELECTION GUIDE
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+  Explore structure          →  list_files (recursive=true)
+  Find exact text in code    →  search_code  (faster than grep for source)
+
+  Find files by name         →  run_bash with find / Get-ChildItem
+  Targeted code replacement  →  edit_file  ← PREFERRED for existing files
+  Create a new file          →  write_file
+  Run tests / linter / build →  run_bash  (pytest, tsc, eslint, cargo test…)
+  Check uncommitted changes  →  git_status
+  Commit completed work      →  git_commit  (surgical staging, not git add .)
+  Remember across sessions   →  write_file → .devpilot/memory.md
+
+{shell_rules}
+
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+COMMUNICATION
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+
+- Be concise. Developers do not want paragraphs of fluff.
+- Before calling a tool, state in one sentence what you are about to do and why.
+- After completing a task (including verification), give a brief summary of
+  what changed and what was verified.
+- If a task is ambiguous, ask one clarifying question before acting.
+{context_section}"""

agent/setup_wizard.py

@@ -0,0 +1,309 @@
+"""
+agent/setup_wizard.py
+──────────────────────
+First-run setup wizard for DevPilot.
+
+Runs when no API key is detected. Guides the user through:
+  1. Choosing a provider (Anthropic, OpenAI, or custom compatible)
+  2. Entering their API key
+  3. Optionally choosing a model
+  4. Saving everything to a .env file in the current directory
+
+Completely non-interactive when running in CI (no TTY).
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+from pathlib import Path
+
+from rich.console import Console
+from rich.panel import Panel
+from rich.prompt import Confirm, Prompt
+from rich.table import Table
+
+console = Console()
+
+# ── Model lists ───────────────────────────────────────────────────────────────
+
+_ANTHROPIC_MODELS = [
+    ("claude-opus-4-5-20251101",   "Most capable — best for complex tasks"),
+    ("claude-sonnet-4-5-20251101", "Balanced — fast and capable"),
+    ("claude-haiku-4-5-20251101",  "Fastest — best for simple tasks"),
+    ("claude-3-7-sonnet-20250219", "Extended thinking support"),
+]
+
+_OPENAI_MODELS = [
+    ("gpt-4o",      "Latest GPT-4o — best for coding"),
+    ("gpt-4o-mini", "Fast and cheap — good for simple tasks"),
+    ("o3",          "Most powerful reasoning model"),
+    ("o4-mini",     "Fast reasoning — great for code"),
+]
+
+# OpenAI-compatible third-party providers
+# (base_url, display_name, key_env_var, key_url, models)
+_COMPATIBLE_PROVIDERS = {
+    "1": (
+        "Groq",
+        "https://console.groq.com/keys",
+        "GROQ_API_KEY",
+        "https://api.groq.com/openai/v1",
+        [
+            ("llama-3.3-70b-versatile", "Best Groq model — fast & capable"),
+            ("llama-3.1-8b-instant",    "Ultra-fast, lightweight"),
+            ("mixtral-8x7b-32768",      "Large context window"),
+            ("gemma2-9b-it",            "Google Gemma 2 — fast"),
+        ],
+    ),
+    "2": (
+        "Together AI",
+        "https://api.together.xyz/settings/api-keys",
+        "TOGETHER_API_KEY",
+        "https://api.together.xyz/v1",
+        [
+            ("meta-llama/Llama-3-70b-chat-hf", "Llama 3 70B — best quality"),
+            ("meta-llama/Llama-3-8b-chat-hf",  "Llama 3 8B — faster"),
+            ("mistralai/Mixtral-8x7B-v0.1",    "Mixtral — large context"),
+        ],
+    ),
+    "3": (
+        "Mistral AI",
+        "https://console.mistral.ai/api-keys/",
+        "MISTRAL_API_KEY",
+        "https://api.mistral.ai/v1",
+        [
+            ("mistral-large-latest", "Most capable Mistral model"),
+            ("mistral-small-latest", "Fast and affordable"),
+            ("codestral-latest",     "Optimised for code"),
+        ],
+    ),
+    "4": (
+        "Ollama (local)",
+        "https://ollama.com/library",
+        "OLLAMA_API_KEY",   # Ollama doesn't need a real key
+        "http://localhost:11434/v1",
+        [
+            ("qwen2.5-coder:7b",  "Best local coding model (recommended)"),
+            ("codellama:13b",     "Meta CodeLlama 13B"),
+            ("deepseek-coder:6b", "DeepSeek Coder 6B"),
+            ("llama3.2:3b",       "Llama 3.2 3B — very fast"),
+        ],
+    ),
+    "5": (
+        "Other (custom)",
+        "",
+        "",
+        "",
+        [],
+    ),
+}
+
+
+# ── Helpers ───────────────────────────────────────────────────────────────────
+
+def _is_interactive() -> bool:
+    return sys.stdin.isatty() and sys.stdout.isatty()
+
+
+def _write_env_file(
+    path: Path,
+    updates: dict[str, str],
+) -> None:
+    """Write or update .env file preserving existing entries."""
+    existing: dict[str, str] = {}
+    if path.exists():
+        for line in path.read_text(encoding="utf-8").splitlines():
+            line = line.strip()
+            if line and not line.startswith("#") and "=" in line:
+                k, _, v = line.partition("=")
+                existing[k.strip()] = v.strip()
+
+    existing.update(updates)
+
+    lines = [
+        "# DevPilot configuration — generated by setup wizard",
+        "# Add this file to .gitignore — never commit API keys!\n",
+    ]
+    for k, v in existing.items():
+        lines.append(f"{k}={v}")
+
+    path.write_text("\n".join(lines) + "\n", encoding="utf-8")
+
+
+def _pick_model(models: list[tuple[str, str]], allow_custom: bool = False) -> str:
+    """Show a model selection table and return the chosen model string."""
+    table = Table(show_header=False, box=None, padding=(0, 2))
+    for i, (name, desc) in enumerate(models, 1):
+        table.add_row(
+            f"[cyan]{i}[/cyan]",
+            f"[bold]{name}[/bold]",
+            f"[dim]{desc}[/dim]",
+        )
+    if allow_custom:
+        table.add_row(
+            f"[cyan]{len(models)+1}[/cyan]",
+            "[bold]Other[/bold]",
+            "[dim]Enter a custom model name[/dim]",
+        )
+    console.print(table)
+
+    choices = [str(i) for i in range(1, len(models) + (2 if allow_custom else 1))]
+    choice = Prompt.ask("\nModel", choices=choices, default="1")
+
+    if allow_custom and choice == str(len(models) + 1):
+        return Prompt.ask("Enter model name").strip()
+
+    return models[int(choice) - 1][0]
+
+
+# ── Main wizard ───────────────────────────────────────────────────────────────
+
+def run_setup_wizard(env_path: Path | None = None) -> bool:
+    """
+    Run the interactive first-run setup wizard.
+    Returns True if setup completed, False if skipped/failed.
+    """
+    if not _is_interactive():
+        return False
+
+    env_path = env_path or Path(".env")
+
+    console.print(Panel(
+        "[bold cyan]Welcome to DevPilot! 🚀[/bold cyan]\n\n"
+        "No API key was found. Let's set up your configuration.\n"
+        "This will create a [bold].env[/bold] file in your current directory.",
+        border_style="cyan",
+        expand=False,
+    ))
+
+    # ── Step 1: Choose provider ───────────────────────────────────────────────
+    console.print("\n[bold]Step 1 of 3 — Choose your AI provider[/bold]\n")
+    console.print("  [cyan]1[/cyan]  Anthropic      (Claude — recommended)")
+    console.print("  [cyan]2[/cyan]  OpenAI         (GPT-4o, o3, o4-mini)")
+    console.print("  [cyan]3[/cyan]  Groq           (Llama 3, Mixtral — very fast, free tier)")
+    console.print("  [cyan]4[/cyan]  Together AI    (Llama 3, Mixtral)")
+    console.print("  [cyan]5[/cyan]  Mistral AI     (Mistral, Codestral)")
+    console.print("  [cyan]6[/cyan]  Ollama         (local models, no API key needed)")
+    console.print("  [cyan]7[/cyan]  Other          (any OpenAI-compatible endpoint)")
+
+    choice = Prompt.ask("\nProvider", choices=["1","2","3","4","5","6","7"], default="1")
+
+    env_updates: dict[str, str] = {}
+
+    # ── Anthropic ─────────────────────────────────────────────────────────────
+    if choice == "1":
+        console.print("\n[bold]Step 2 of 3 — Enter your Anthropic API key[/bold]")
+        console.print("  Get one at: [link=https://console.anthropic.com/]https://console.anthropic.com/[/link]")
+        api_key = Prompt.ask("\nANTHROPIC_API_KEY", password=True).strip()
+        if not api_key:
+            console.print("[red]API key cannot be empty.[/red]")
+            return False
+        if not api_key.startswith("sk-ant-"):
+            console.print("[yellow]⚠  Key doesn't look like a typical Anthropic key (sk-ant-...). Saving anyway.[/yellow]")
+
+        console.print("\n[bold]Step 3 of 3 — Choose a model[/bold]\n")
+        model = _pick_model(_ANTHROPIC_MODELS)
+
+        env_updates = {
+            "ANTHROPIC_API_KEY":  api_key,
+            "DEVPILOT_PROVIDER":  "anthropic",
+            "DEVPILOT_MODEL":     model,
+        }
+
+    # ── OpenAI ────────────────────────────────────────────────────────────────
+    elif choice == "2":
+        console.print("\n[bold]Step 2 of 3 — Enter your OpenAI API key[/bold]")
+        console.print("  Get one at: [link=https://platform.openai.com/api-keys]https://platform.openai.com/api-keys[/link]")
+        api_key = Prompt.ask("\nOPENAI_API_KEY", password=True).strip()
+        if not api_key:
+            console.print("[red]API key cannot be empty.[/red]")
+            return False
+        if not api_key.startswith("sk-"):
+            console.print("[yellow]⚠  Key doesn't look like a typical OpenAI key (sk-...). Saving anyway.[/yellow]")
+
+        console.print("\n[bold]Step 3 of 3 — Choose a model[/bold]\n")
+        model = _pick_model(_OPENAI_MODELS, allow_custom=True)
+
+        env_updates = {
+            "OPENAI_API_KEY":    api_key,
+            "DEVPILOT_PROVIDER": "openai",
+            "DEVPILOT_MODEL":    model,
+        }
+
+    # ── Compatible providers (Groq, Together, Mistral, Ollama) ───────────────
+    elif choice in ("3", "4", "5", "6"):
+        compat_key = str(int(choice) - 2)  # maps 3→1, 4→2, 5→3, 6→4
+        name, key_url, key_env, base_url, models = _COMPATIBLE_PROVIDERS[compat_key]
+
+        console.print(f"\n[bold]Step 2 of 3 — Enter your {name} API key[/bold]")
+
+        if choice == "6":  # Ollama — no real key needed
+            console.print("  [dim]Ollama runs locally — no API key required.[/dim]")
+            console.print("  Make sure Ollama is running: [bold]ollama serve[/bold]")
+            api_key = "ollama"
+        else:
+            console.print(f"  Get one at: [link={key_url}]{key_url}[/link]")
+            api_key = Prompt.ask(f"\n{key_env}", password=True).strip()
+            if not api_key:
+                console.print("[red]API key cannot be empty.[/red]")
+                return False
+
+        console.print(f"\n[bold]Step 3 of 3 — Choose a model[/bold]\n")
+        model = _pick_model(models, allow_custom=True)
+
+        env_updates = {
+            key_env:                api_key,
+            "OPENAI_API_KEY":       api_key,  # DevPilot reads OPENAI_API_KEY for openai provider
+            "DEVPILOT_PROVIDER":    "openai",
+            "DEVPILOT_MODEL":       model,
+            "DEVPILOT_BASE_URL":    base_url,
+        }
+
+    # ── Custom endpoint ───────────────────────────────────────────────────────
+    elif choice == "7":
+        console.print("\n[bold]Step 2 of 3 — Custom OpenAI-compatible endpoint[/bold]")
+        base_url  = Prompt.ask("Base URL (e.g. https://api.example.com/v1)").strip()
+        api_key   = Prompt.ask("API key", password=True).strip()
+        model     = Prompt.ask("Model name (e.g. llama-3-70b)").strip()
+
+        if not base_url or not model:
+            console.print("[red]Base URL and model name are required.[/red]")
+            return False
+
+        env_updates = {
+            "OPENAI_API_KEY":    api_key or "none",
+            "DEVPILOT_PROVIDER": "openai",
+            "DEVPILOT_MODEL":    model,
+            "DEVPILOT_BASE_URL": base_url,
+        }
+
+    # ── Save .env ─────────────────────────────────────────────────────────────
+    try:
+        _write_env_file(env_path, env_updates)
+    except OSError as e:
+        console.print(f"[red]Failed to write .env file: {e}[/red]")
+        return False
+
+    # Inject into current process immediately
+    for k, v in env_updates.items():
+        os.environ[k] = v
+
+    provider_display = env_updates.get("DEVPILOT_PROVIDER", "openai")
+    model_display    = env_updates.get("DEVPILOT_MODEL", "")
+    base_url_display = env_updates.get("DEVPILOT_BASE_URL", "")
+
+    summary = (
+        f"[bold green]✓ Setup complete![/bold green]\n\n"
+        f"  Provider : [cyan]{provider_display}[/cyan]\n"
+        f"  Model    : [cyan]{model_display}[/cyan]\n"
+    )
+    if base_url_display:
+        summary += f"  Base URL : [cyan]{base_url_display}[/cyan]\n"
+    summary += (
+        f"  Saved to : [cyan]{env_path.resolve()}[/cyan]\n\n"
+        "[dim]Add .env to your .gitignore — never commit API keys![/dim]"
+    )
+
+    console.print(Panel(summary, border_style="green", expand=False))
+    return True

agent/tools/__init__.py

@@ -0,0 +1,15 @@
+"""
+agent/tools package.
+Contains all built-in tools and the ToolRegistry.
+"""
+
+from agent.tools.base import ToolResult, ToolSchema, BaseTool
+from agent.tools.registry import ToolRegistry, PermissionGuard
+
+__all__ = [
+    "ToolResult",
+    "ToolSchema",
+    "BaseTool",
+    "ToolRegistry",
+    "PermissionGuard",
+]

agent/tools/a2a.py

@@ -0,0 +1,56 @@
+"""
+agent/tools/a2a.py
+──────────────────
+A2A delegation tool.
+"""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
+
+from agent.a2a_client import delegate_task_to_peer
+from agent.tools.base import BaseTool, ToolResult, ToolSchema
+
+if TYPE_CHECKING:
+    from agent.config import Config
+
+
+class A2ATool(BaseTool):
+    """Delegate tasks to peer agents."""
+
+    def __init__(self, config: "Config") -> None:
+        self._config = config
+
+    @property
+    def schema(self) -> ToolSchema:
+        return ToolSchema(
+            name="a2a_delegate_task",
+            description=(
+                "Delegate a subtask to an external A2A (Agent-to-Agent) peer. "
+                "Use this when you need help from a specialist agent or another node. "
+                "Provide the base URL of the peer agent and the task prompt."
+            ),
+            parameters={
+                "type": "object",
+                "properties": {
+                    "peer_url": {
+                        "type": "string",
+                        "description": "Base URL of the peer agent (e.g., http://localhost:8001)"
+                    },
+                    "prompt": {
+                        "type": "string",
+                        "description": "The coding task or instruction to delegate."
+                    },
+                    "token": {
+                        "type": "string",
+                        "description": "Optional Bearer token if the peer requires authentication."
+                    }
+                },
+                "required": ["peer_url", "prompt"]
+            },
+            required=["peer_url", "prompt"],
+            sprint="Sprint 4",
+        )
+
+    async def execute(self, peer_url: str, prompt: str, token: str | None = None) -> ToolResult:  # type: ignore[override]
+        return await delegate_task_to_peer(peer_url, prompt, token)

agent/tools/base.py

@@ -0,0 +1,52 @@
+"""
+agent/tools/base.py
+───────────────────
+Base interfaces for DevPilot tools.
+"""
+
+from __future__ import annotations
+
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import Any
+
+@dataclass
+class ToolResult:
+    """Returned by every tool executor."""
+    output: str       # String content to send back to the model
+    is_error: bool    # True if the tool raised an error or was cancelled
+
+
+@dataclass
+class ToolSchema:
+    """Portable tool schema — provider-agnostic."""
+    name: str
+    description: str
+    parameters: dict[str, Any]           # JSON Schema object for input
+    required: list[str] = field(default_factory=list)
+    is_destructive: bool = False         # If True, permission guard prompts
+    sprint: str = "Sprint 1"            # Implemented in which sprint
+
+
+class BaseTool(ABC):
+    """Abstract base class for all DevPilot tools."""
+
+    @property
+    @abstractmethod
+    def schema(self) -> ToolSchema:
+        """Return the tool's JSON schema for the model."""
+
+    @abstractmethod
+    async def execute(self, **kwargs: Any) -> ToolResult:
+        """
+        Execute the tool with the arguments the model provided.
+        Returns a ToolResult containing output and error status.
+        """
+
+    @property
+    def name(self) -> str:
+        return self.schema.name
+
+    @property
+    def is_destructive(self) -> bool:
+        return self.schema.is_destructive