CascadingDev/automation/runner.py

"""
automation.runner
=================

Entry point invoked by the pre-commit hook to evaluate `.ai-rules.yml`
instructions.  The runner is intentionally thin: it inspects staged files,
looks up the matching rule/output definitions, merges instruction strings and
delegates execution to `automation.patcher.generate_output`, which handles the
heavy lifting (prompt composition, AI invocation, patch application).
"""
from __future__ import annotations

import argparse
import sys
from pathlib import Path
from typing import Dict, Iterable

from automation.config import RulesConfig
from automation.patcher import ModelConfig, generate_output, run


def get_staged_files(repo_root: Path) -> list[Path]:
    """
    Return staged (added/modified) paths relative to the repository root.
    """
    # We only care about what is in the index; the working tree may include
    # experiments the developer does not intend to commit. `--diff-filter=AM`
    # narrows the list to new or modified files.
    result = run(
        ["git", "diff", "--cached", "--name-only", "--diff-filter=AM"],
        cwd=repo_root,
        check=False,
    )
    return [Path(line.strip()) for line in result.stdout.splitlines() if line.strip()]


def merge_instructions(source_instr: str, output_instr: str, append_instr: str) -> str:
    """
    Combine source-level, output-level, and append instructions into a single prompt.
    """
    final = output_instr.strip() if output_instr else source_instr.strip()
    if not final:
        final = source_instr.strip()
    append_instr = append_instr.strip()
    if append_instr:
        prefix = (final + "\n\n") if final else ""
        final = f"{prefix}Additional requirements for this output location:\n{append_instr}"
    return final.strip()  # Final, human-readable instruction block handed to the AI


def process(repo_root: Path, rules: RulesConfig, model: ModelConfig) -> int:
    """
    Walk staged files, resolve matching outputs, and invoke the patcher for each.
    """
    # 1) Gather the staged file list (Git index only).
    staged_files = get_staged_files(repo_root)
    if not staged_files:
        return 0

    # 2) For each staged file, look up the matching rule and iterate outputs.
    for src_rel in staged_files:
        # Find the most specific rule (nearest .ai-rules.yml wins).
        rule_name = rules.get_rule_name(src_rel)
        if not rule_name:
            continue

        rule_config = rules.cascade_for(src_rel, rule_name)
        outputs: Dict[str, Dict] = rule_config.get("outputs") or {}
        source_instruction = rule_config.get("instruction", "")

        for output_name, output_cfg in outputs.items():
            if not isinstance(output_cfg, dict):
                continue
            if str(output_cfg.get("enabled", "true")).lower() == "false":
                continue

            path_template = output_cfg.get("path")
            if not path_template:
                continue

            rendered_path = rules.resolve_template(path_template, src_rel)
            try:
                output_rel = rules.normalize_repo_rel(rendered_path)
            except ValueError:
                print(f"[runner] skipping {output_name}: unsafe path {rendered_path}", file=sys.stderr)
                continue

            # Build the instruction set for this output. Output-specific text
            # overrides the rule-level text, and we keep the source version as a
            # fallback.
            instruction = output_cfg.get("instruction", "") or source_instruction
            append = output_cfg.get("instruction_append", "")
            model_hint = rule_config.get("model_hint", "")

            output_type = output_cfg.get("output_type")
            if output_type:
                extra = rules.cascade_for(output_rel, output_type)
                instruction = extra.get("instruction", instruction)
                append = extra.get("instruction_append", append)
                # Output type can also override model hint
                if "model_hint" in extra:
                    model_hint = extra["model_hint"]

            final_instruction = merge_instructions(source_instruction, instruction, append)

            # 3) Ask the patcher to build a diff with the assembled instruction.
            try:
                print(f"[runner] generating {output_rel.as_posix()} from {src_rel.as_posix()}")
                generate_output(
                    repo_root=repo_root,
                    rules=rules,
                    model=model,
                    source_rel=src_rel,
                    output_rel=output_rel,
                    instruction=final_instruction,
                    model_hint=model_hint,
                )
            except Exception as exc:  # pragma: no cover - defensive
                print(f"[runner] error generating {output_rel}: {exc}", file=sys.stderr)

    return 0


def main(argv: list[str] | None = None) -> int:
    """
    CLI entry point used by the pre-commit hook.
    """
    # Parse command-line options (only --model override today).
    parser = argparse.ArgumentParser(description="CascadingDev AI runner")
    parser.add_argument("--model", help="Override AI command (default from env)")
    args = parser.parse_args(argv)

    # Load the nearest .ai-rules.yml (fail quietly if missing).
    repo_root = Path.cwd().resolve()
    try:
        rules = RulesConfig.load(repo_root)
    except FileNotFoundError:
        print("[runner] .ai-rules.yml not found; skipping")
        return 0

    # Instantiate the model config and delegate to the processing pipeline.
    model = ModelConfig.from_sources(repo_root, args.model)
    return process(repo_root, rules, model)


if __name__ == "__main__":
    sys.exit(main())