Upload 4 files

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+marl/core.cpython-312-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text
+marl/proxy.cpython-312-x86_64-linux-gnu.so filter=lfs diff=lfs merge=lfs -text

marl/__init__.py

@@ -0,0 +1,29 @@
+"""
+MARL — Model-Agnostic Runtime Middleware for LLMs
+═══════════════════════════════════════════════════
+Apply the 5-stage multi-agent pipeline (S1→S2→S3→S4→S5) to ANY LLM
+to systematically improve reasoning, self-correction, and reliability.
+
+Usage:
+    from marl import Marl
+
+    # OpenAI
+    marl = Marl.from_openai(api_key="sk-...")
+    result = marl.run("Your complex question here")
+
+    # Any custom LLM
+    marl = Marl(call_fn=my_llm_function)
+    result = marl.run("Your question")
+
+    # As OpenAI-compatible proxy
+    marl.serve(port=8080)  # localhost:8080/v1/chat/completions
+
+Author: Ginigen AI
+License: Apache 2.0
+"""
+
+from .core import Marl, MarlResult, MarlConfig
+from .proxy import MarlProxy
+
+__version__ = "1.0.0"
+__all__ = ["Marl", "MarlResult", "MarlConfig", "MarlProxy"]

marl/__main__.py

@@ -0,0 +1,116 @@
+"""
+MARL CLI — Command-line interface
+
+Usage:
+    # Run as proxy
+    python -m marl proxy --port 8080 --backend openai --model gpt-5.2
+
+    # Run single query
+    python -m marl run "Your question here" --backend ollama --model llama3.1
+
+    # Test connection
+    python -m marl test --backend openai
+"""
+
+import argparse
+import os
+import sys
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        prog="marl",
+        description="🌀 MARL — Model-Agnostic Runtime Middleware for LLMs"
+    )
+    sub = parser.add_subparsers(dest="command")
+
+    # ── proxy ──
+    p_proxy = sub.add_parser("proxy", help="Start OpenAI-compatible proxy server")
+    p_proxy.add_argument("--port", type=int, default=8080)
+    p_proxy.add_argument("--host", default="0.0.0.0")
+    p_proxy.add_argument("--backend", default="openai",
+                         choices=["openai", "anthropic", "ollama", "friendli", "custom"])
+    p_proxy.add_argument("--model", default=None)
+    p_proxy.add_argument("--api-key", default=None)
+    p_proxy.add_argument("--base-url", default=None, help="For custom backend")
+
+    # ── run ──
+    p_run = sub.add_parser("run", help="Run single MARL query")
+    p_run.add_argument("prompt", help="The question/task")
+    p_run.add_argument("--backend", default="openai",
+                       choices=["openai", "anthropic", "ollama", "friendli", "custom"])
+    p_run.add_argument("--model", default=None)
+    p_run.add_argument("--api-key", default=None)
+    p_run.add_argument("--base-url", default=None)
+    p_run.add_argument("--trace", action="store_true", help="Show full agent trace")
+
+    # ── test ──
+    p_test = sub.add_parser("test", help="Test backend connection")
+    p_test.add_argument("--backend", default="openai",
+                        choices=["openai", "anthropic", "ollama", "friendli", "custom"])
+    p_test.add_argument("--model", default=None)
+    p_test.add_argument("--api-key", default=None)
+    p_test.add_argument("--base-url", default=None)
+
+    args = parser.parse_args()
+
+    if not args.command:
+        parser.print_help()
+        return
+
+    from .core import Marl, MarlConfig
+
+    marl = _build_marl(args)
+
+    if args.command == "proxy":
+        marl.serve(host=args.host, port=args.port)
+
+    elif args.command == "run":
+        config = MarlConfig(include_trace=args.trace)
+        marl.config = config
+        print("🌀 MARL running S1→S2→S3→S4→S5 pipeline...\n")
+        result = marl.run(args.prompt)
+        if args.trace:
+            print(result.full_output)
+        else:
+            print(result.answer)
+        print(f"\n⏱️ {result.elapsed:.1f}s | Fixes: {len(result.fixes)}")
+
+    elif args.command == "test":
+        print("🔍 Testing backend connection...")
+        try:
+            resp = marl.call_fn("Say OK", "", 10, 0)
+            if resp and not resp.startswith("[ERROR"):
+                print(f"✅ Connected! Response: {resp[:50]}")
+            else:
+                print(f"❌ Error: {resp}")
+        except Exception as e:
+            print(f"❌ Failed: {e}")
+
+
+def _build_marl(args) -> "Marl":
+    from .core import Marl
+    backend = args.backend
+    api_key = args.api_key or os.getenv("OPENAI_API_KEY") or os.getenv("ANTHROPIC_API_KEY") or ""
+    model = args.model
+
+    if backend == "openai":
+        return Marl.from_openai(api_key=api_key, model=model or "gpt-5.2")
+    elif backend == "anthropic":
+        key = args.api_key or os.getenv("ANTHROPIC_API_KEY", "")
+        return Marl.from_anthropic(api_key=key, model=model or "claude-sonnet-4-20250514")
+    elif backend == "ollama":
+        return Marl.from_ollama(model=model or "llama3.1")
+    elif backend == "friendli":
+        key = args.api_key or os.getenv("FRIENDLI_TOKEN", "")
+        return Marl.from_friendli(token=key, model=model or "deppfs281rgffnk")
+    elif backend == "custom":
+        base_url = args.base_url or "http://localhost:8000/v1"
+        return Marl.from_openai_compatible(base_url=base_url, api_key=api_key,
+                                            model=model or "default")
+    else:
+        raise ValueError(f"Unknown backend: {backend}")
+
+
+if __name__ == "__main__":
+    main()

marl/core.cpython-312-x86_64-linux-gnu.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4961041acda79d7c394b215734a9d93510cd84aa397431062b075ddd6ad60291
+size 1478112

marl/proxy.cpython-312-x86_64-linux-gnu.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b597e6159ccc821ac7a8b8bd6c8a1b4209175742d5aae38fc1fc7361282b4199
+size 514720