init

post_llama.py

@@ -0,0 +1,149 @@
+#!/usr/bin/env python3
+"""
+Post a message with an attached file (contents embedded in JSON) to an
+OpenAI‑compatible API and echo the response.
+
+Defaults:
+    • Server: https://llama-cpp.reeselink.com/v1/chat/completions
+    • Model : gpt-oss-120b
+    • No API key required (but can be supplied via --api-key)
+"""
+
+import argparse
+import json
+import sys
+from pathlib import Path
+
+import requests
+
+# ----------------------------------------------------------------------
+# Helper functions
+# ----------------------------------------------------------------------
+def load_file_text(path: Path) -> str:
+    """Read the whole file as UTF‑8 text (fallback to binary → base64 if needed)."""
+    try:
+        return path.read_text(encoding="utf-8")
+    except UnicodeDecodeError:
+        # For non‑text files we fall back to a base64 representation.
+        import base64
+        return base64.b64encode(path.read_bytes()).decode("ascii")
+
+
+def build_payload(
+    model: str,
+    user_message: str,
+    file_content: str,
+    file_name: str,
+) -> dict:
+    """
+    Construct the JSON body expected by the OpenAI chat completions endpoint.
+
+    The “file attachment” is modeled as an extra user‑role message that contains
+    the file name and its contents.  This mirrors the way many front‑ends embed
+    files in the conversation history.
+    """
+    # Primary user message (the prompt you want the model to answer)
+    messages = [{"role": "user", "content": user_message}]
+
+    # Add a second message that represents the attached file.
+    # The format is arbitrary – you can change it to suit your downstream model.
+    file_message = (
+        f"[Attached file: {file_name}]\n"
+        f"{file_content}"
+    )
+    messages.append({"role": "user", "content": file_message})
+
+    return {
+        "model": model,
+        "messages": messages,
+        # Optional parameters – feel free to expose them as CLI args later.
+        "temperature": 0.7,
+        "max_tokens": 1024,
+    }
+
+
+def post_chat_completion(
+    endpoint: str,
+    payload: dict,
+    api_key: str | None = None,
+) -> requests.Response:
+    """POST the JSON payload to the API and return the raw response."""
+    headers = {"Content-Type": "application/json"}
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
+
+    response = requests.post(endpoint, headers=headers, json=payload, timeout=300)
+    response.raise_for_status()  # raise an exception for HTTP errors
+    return response
+
+
+# ----------------------------------------------------------------------
+# CLI entry point
+# ----------------------------------------------------------------------
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Send a message + file contents to an OpenAI‑compatible API."
+    )
+    parser.add_argument(
+        "file",
+        type=Path,
+        help="Path to the file whose contents will be sent in the payload.",
+    )
+    parser.add_argument(
+        "-m",
+        "--message",
+        default="Please analyse the attached file.",
+        help="User message to send alongside the file.",
+    )
+    parser.add_argument(
+        "--model",
+        default="gpt-oss-120b",
+        help="Model name to request (default: %(default)s).",
+    )
+    parser.add_argument(
+        "--api-key",
+        default=None,
+        help="Optional API key; omit if the server does not require authentication.",
+    )
+    parser.add_argument(
+        "--url",
+        default="https://llama-cpp.reeselink.com/v1/chat/completions",
+        help="Base URL of the OpenAI‑compatible endpoint (default: %(default)s).",
+    )
+    args = parser.parse_args()
+
+    # ------------------------------------------------------------------
+    # Read file and build request
+    # ------------------------------------------------------------------
+    if not args.file.is_file():
+        sys.exit(f"Error: '{args.file}' does not exist or is not a regular file.")
+
+    file_content = load_file_text(args.file)
+    payload = build_payload(
+        model=args.model,
+        user_message=args.message,
+        file_content=file_content,
+        file_name=args.file.name,
+    )
+
+    # ------------------------------------------------------------------
+    # Send request
+    # ------------------------------------------------------------------
+    try:
+        resp = post_chat_completion(endpoint=args.url, payload=payload, api_key=args.api_key)
+    except requests.RequestException as exc:
+        sys.exit(f"Request failed: {exc}")
+
+    # ------------------------------------------------------------------
+    # Echo the response (pretty‑print JSON if possible)
+    # ------------------------------------------------------------------
+    try:
+        json_resp = resp.json()
+        print(json.dumps(json_resp["choices"][0]["message"]["content"], indent=2, ensure_ascii=False))
+    except ValueError:
+        # Not JSON – just dump raw text
+        print(resp.text)
+
+
+if __name__ == "__main__":
+    main()