Spaces:
Sleeping
Sleeping
| """CLI entry point: ask the agent a question from the terminal. | |
| Usage: | |
| python scripts/ask.py --config config.yaml "What was FY revenue?" | |
| It builds the AgenticRAG pipeline from a config file, runs one question through it, and | |
| pretty-prints the answer, the citations, and the step/usage stats. | |
| """ | |
| import argparse # parses command-line flags/arguments | |
| import sys, pathlib # used to put the project's src/ directory on the import path | |
| # The package lives under <project>/src. This line inserts that folder at the front of | |
| # sys.path so `import agentic_rag...` works even when running the script directly (no install). | |
| # __file__ -> this script -> .parent (scripts/) -> .parent (project root) -> /src | |
| sys.path.insert(0, str(pathlib.Path(__file__).resolve().parent.parent / "src")) | |
| from agentic_rag.agent import AgenticRAG | |
| def main(): | |
| """Parse args, run one question, print the answer + citations + stats.""" | |
| # --- Define the CLI: an optional --config and one-or-more positional words for the question. | |
| ap = argparse.ArgumentParser() | |
| ap.add_argument("--config", default="config.yaml") # path to the YAML config | |
| ap.add_argument("question", nargs="+") # question words (joined with spaces below) | |
| args = ap.parse_args() | |
| # --- Build the pipeline and answer the question. | |
| rag = AgenticRAG.from_config(args.config) | |
| result = rag.answer(" ".join(args.question)) # rejoin the words into one query string | |
| # --- Print the answer. | |
| print("\n=== ANSWER ===\n" + result.text) | |
| # --- Print every citation: [source pPage] chunk <id> score <score>. | |
| print(f"\n=== CITATIONS ({len(result.citations)}) ===") | |
| for c in result.citations: | |
| print(f" [{c['source']} p{c['page']}] chunk {c['chunk_id']} score {c['score']}") | |
| # --- Print how many search steps ran and the token usage. | |
| print(f"\nsteps={result.steps} usage={result.usage}") | |
| # Standard Python idiom: only run main() when this file is executed directly | |
| # (python scripts/ask.py ...), not when it's imported by another module. | |
| if __name__ == "__main__": | |
| main() | |