synth-ai 0.2.9.dev0__py3-none-any.whl → 0.2.23.dev3__py3-none-any.whl

examples/qwen_vl/crafter_qwen_vl_agent.py

@@ -0,0 +1,300 @@
+#!/usr/bin/env python3
+"""
+Crafter agent using Qwen-VL models via synth-ai's hosted inference.
+
+This demonstrates vision-language models (Qwen3-VL family) playing Crafter
+with image observations. The CrafterPolicy automatically detects vision capability
+from the model name and includes base64-encoded PNG frames in the prompt.
+
+Requirements:
+  - `SYNTH_API_KEY` environment variable (for synth-ai hosted inference)
+  - synth-ai package with Crafter task app dependencies
+
+Usage:
+  uv run python examples/qwen_vl/crafter_qwen_vl_agent.py \
+      --model Qwen/Qwen3-VL-8B-Instruct --seeds 10 --steps 20
+"""
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import base64
+import json
+import os
+from contextlib import suppress
+from pathlib import Path
+from typing import Any
+from uuid import uuid4
+
+from examples.task_apps.crafter.task_app.synth_envs_hosted.envs.crafter.environment import (
+    CrafterEnvironmentWrapper,
+)
+from examples.task_apps.crafter.task_app.synth_envs_hosted.envs.crafter.policy import CrafterPolicy
+from synth_ai.environments.examples.crafter_classic.environment import CrafterClassicEnvironment
+from synth_ai.environments.examples.crafter_classic.taskset import (
+    CrafterTaskInstance,
+    CrafterTaskInstanceMetadata,
+)
+from synth_ai.environments.tasks.core import Impetus, Intent
+
+# Import synth-ai inference client
+try:
+    from synth_ai.inference.client import InferenceClient
+except ImportError:
+    print("Error: synth-ai inference client not found. Make sure synth-ai is installed.")
+    raise
+
+DEFAULT_OUTPUT = Path("examples/qwen_vl/temp")
+FRAME_SUBDIR = "qwen_vl_frames"
+
+
+def _default_backend_base_url() -> str:
+    raw = os.getenv("BACKEND_BASE_URL", "https://agent-learning.onrender.com/api").strip()
+    return raw if raw.endswith("/api") else f"{raw}/api"
+
+
+class EpisodeResult:
+    def __init__(self, seed: int) -> None:
+        self.seed = seed
+        self.steps_taken: int = 0
+        self.achievements: set[str] = set()
+        self.total_reward: float = 0.0
+        self.tool_calls: int = 0
+
+    def record_observation(self, observation: dict[str, Any]) -> None:
+        obs = observation.get("observation") if isinstance(observation, dict) else None
+        if not isinstance(obs, dict):
+            return
+        ach = obs.get("achievements_status")
+        if isinstance(ach, dict):
+            for name, unlocked in ach.items():
+                if unlocked:
+                    self.achievements.add(str(name))
+        reward = obs.get("reward_last_step")
+        if isinstance(reward, int | float):
+            self.total_reward += float(reward)
+
+
+def _ensure_synth_client() -> InferenceClient:
+    """Initialize synth-ai inference client."""
+    api_key = os.getenv("SYNTH_API_KEY")
+    if not api_key:
+        raise RuntimeError(
+            "SYNTH_API_KEY must be set for synth-ai hosted inference. "
+            "Get your key from https://synth-ai.com"
+        )
+    base_url = os.getenv("SYNTH_BASE_URL", _default_backend_base_url())
+    return InferenceClient(base_url=base_url, api_key=api_key)
+
+
+def _build_task_instance(seed: int) -> CrafterTaskInstance:
+    """Create a Crafter task instance with specified seed."""
+    impetus = Impetus(instructions="Explore, survive, and unlock achievements.")
+    intent = Intent(
+        rubric={"goal": "Maximise Crafter achievements."},
+        gold_trajectories=None,
+        gold_state_diff={},
+    )
+    metadata = CrafterTaskInstanceMetadata(
+        difficulty="custom",
+        seed=seed,
+        num_trees_radius=0,
+        num_cows_radius=0,
+        num_hostiles_radius=0,
+    )
+    instance = CrafterTaskInstance(
+        id=uuid4(),
+        impetus=impetus,
+        intent=intent,
+        metadata=metadata,
+        is_reproducible=True,
+        initial_engine_snapshot=None,
+    )
+    setattr(instance, "config", {"seed": seed, "length": 256, "area": [64, 64]})
+    return instance
+
+
+def _decode_and_save_image(observation: dict[str, Any], path: Path) -> None:
+    """Extract and save PNG frame from observation."""
+    obs = observation.get("observation") if isinstance(observation, dict) else None
+    if not isinstance(obs, dict):
+        return
+    base64_data = obs.get("observation_image_base64")
+    if not isinstance(base64_data, str) or not base64_data:
+        return
+    path.parent.mkdir(parents=True, exist_ok=True)
+    with suppress(Exception):
+        path.write_bytes(base64.b64decode(base64_data))
+
+
+async def _run_episode(
+    *,
+    seed: int,
+    client: InferenceClient,
+    model: str,
+    max_steps: int,
+    output_dir: Path,
+    temperature: float,
+) -> EpisodeResult:
+    """Run a single Crafter episode with Qwen-VL."""
+    task_instance = _build_task_instance(seed)
+    env = CrafterClassicEnvironment(task_instance)
+    wrapper = CrafterEnvironmentWrapper(env, seed=seed)
+    
+    # Policy will auto-detect vision from model name (qwen-vl and qwen3-vl tokens)
+    policy = CrafterPolicy(inference_url="synth://inference", model=model)
+    await policy.initialize({
+        "use_tools": True,
+        "model": model,
+        "temperature": temperature,
+        "max_tokens": 512,
+    })
+
+    episode_result = EpisodeResult(seed=seed)
+
+    observation_packet = await wrapper.initialize()
+    episode_result.record_observation(observation_packet)
+
+    frames_root = output_dir / FRAME_SUBDIR / f"seed_{seed:04d}"
+    _decode_and_save_image(observation_packet, frames_root / "step_000.png")
+
+    for step_idx in range(max_steps):
+        obs_dict = observation_packet.get("observation")
+        if not isinstance(obs_dict, dict):
+            break
+
+        # Format observation text
+        obs_text = policy._format_observation_for_llm(observation_packet)  # noqa: SLF001
+
+        # Get tool calls from policy (it prepares the inference request internally)
+        tool_calls, meta = await policy.step(
+            observation_text=obs_text,
+            metadata={"raw_observation": observation_packet},
+        )
+        if "inference_request" not in meta:
+            break
+
+        episode_result.steps_taken += 1
+        inference_request = meta["inference_request"]
+
+        # Call synth-ai hosted inference
+        response = await client.create_chat_completion(
+            model=model,
+            messages=inference_request["messages"],
+            temperature=temperature,
+            max_tokens=512,
+            tools=inference_request.get("tools"),
+        )
+
+        # Parse tool calls from response
+        assistant_tool_calls = CrafterPolicy.parse_response_to_tool_calls(
+            response,
+            use_tools=policy.use_tools,
+        )
+        if not assistant_tool_calls:
+            print(
+                f"Seed {seed}: no tool calls returned by model; ending episode early at step {step_idx}."
+            )
+            break
+
+        episode_result.tool_calls += len(assistant_tool_calls)
+
+        # Extract assistant message
+        assistant_message = response.get("choices", [{}])[0].get("message", {})
+        assistant_text = assistant_message.get("content")
+
+        # Execute action in environment
+        env_response = await wrapper.step(assistant_tool_calls)
+        if not isinstance(env_response, dict):
+            raise RuntimeError(
+                f"Unexpected environment response type: {type(env_response)!r}"
+            )
+        episode_result.record_observation(env_response)
+
+        # Update policy history
+        policy._append_assistant_turn(  # noqa: SLF001
+            assistant_text,
+            assistant_tool_calls,
+            env_response,
+        )
+
+        # Save frame
+        frame_path = frames_root / f"step_{step_idx + 1:03d}.png"
+        _decode_and_save_image(env_response, frame_path)
+
+        if env_response.get("done"):
+            break
+        observation_packet = env_response
+
+    await wrapper.terminate()
+    return episode_result
+
+
+async def main() -> None:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--model",
+        default="Qwen/Qwen3-VL-8B-Instruct",
+        help="Qwen-VL model name (e.g., Qwen/Qwen3-VL-2B-Instruct, Qwen/Qwen3-VL-8B-Instruct)",
+    )
+    parser.add_argument("--seeds", type=int, default=10, help="Number of random seeds to evaluate")
+    parser.add_argument("--steps", type=int, default=20, help="Max steps per seed")
+    parser.add_argument("--temperature", type=float, default=0.7, help="Sampling temperature")
+    parser.add_argument(
+        "--output-dir",
+        type=Path,
+        default=DEFAULT_OUTPUT,
+        help=f"Directory for saved frames and summaries (default: {DEFAULT_OUTPUT})",
+    )
+    args = parser.parse_args()
+
+    client = _ensure_synth_client()
+    results: list[EpisodeResult] = []
+
+    seeds = list(range(args.seeds))
+    print(f"Running {len(seeds)} Crafter episodes with model={args.model}")
+    print(f"Using synth-ai hosted inference\n")
+
+    for seed in seeds:
+        result = await _run_episode(
+            seed=seed,
+            client=client,
+            model=args.model,
+            max_steps=args.steps,
+            output_dir=args.output_dir,
+            temperature=args.temperature,
+        )
+        results.append(result)
+        print(
+            f"Seed {seed:02d}: steps={result.steps_taken}, "
+            f"achievements={len(result.achievements)}, "
+            f"tool_calls={result.tool_calls}, reward≈{result.total_reward:.3f}"
+        )
+
+    summary = {
+        "model": args.model,
+        "provider": "synth-ai",
+        "episodes": len(results),
+        "mean_steps": round(
+            sum(res.steps_taken for res in results) / max(len(results), 1), 2
+        ),
+        "mean_achievements": round(
+            sum(len(res.achievements) for res in results) / max(len(results), 1), 2
+        ),
+        "total_tool_calls": sum(res.tool_calls for res in results),
+        "output_dir": str(args.output_dir / FRAME_SUBDIR),
+    }
+
+    args.output_dir.mkdir(parents=True, exist_ok=True)
+    summary_path = args.output_dir / "qwen_vl_summary.json"
+    summary_path.write_text(json.dumps(summary, indent=2), encoding="utf-8")
+
+    print("\nSummary")
+    print("-------")
+    print(json.dumps(summary, indent=2))
+    print(f"\nFrames saved in: {summary['output_dir']}")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

examples/qwen_vl/run_vision_comparison.sh

@@ -0,0 +1,61 @@
+#!/bin/bash
+# Compare Qwen-VL (via synth) vs gpt-5-nano (via OpenAI) on Crafter
+
+set -e
+
+SEEDS=10
+STEPS=20
+OUTPUT_DIR="examples/qwen_vl/temp/comparison"
+
+echo "======================================"
+echo "Vision Model Comparison on Crafter"
+echo "======================================"
+echo ""
+echo "Running $SEEDS episodes, $STEPS steps each"
+echo ""
+
+# Check API keys
+if [ -z "$OPENAI_API_KEY" ]; then
+    echo "Error: OPENAI_API_KEY not set"
+    exit 1
+fi
+
+if [ -z "$SYNTH_API_KEY" ]; then
+    echo "Error: SYNTH_API_KEY not set"
+    exit 1
+fi
+
+# Run gpt-5-nano
+echo "======================================"
+echo "1. Running gpt-5-nano (OpenAI)"
+echo "======================================"
+uv run python examples/qwen_vl/crafter_gpt5nano_agent.py \
+    --model gpt-5-nano \
+    --seeds $SEEDS \
+    --steps $STEPS \
+    --output-dir "$OUTPUT_DIR/gpt5nano"
+
+echo ""
+echo "======================================"
+echo "2. Running Qwen3-VL-8B (synth-ai)"
+echo "======================================"
+uv run python examples/qwen_vl/crafter_qwen_vl_agent.py \
+    --model Qwen/Qwen3-VL-8B-Instruct \
+    --seeds $SEEDS \
+    --steps $STEPS \
+    --output-dir "$OUTPUT_DIR/qwen3vl"
+
+echo ""
+echo "======================================"
+echo "Results Summary"
+echo "======================================"
+echo ""
+echo "gpt-5-nano (OpenAI):"
+cat "$OUTPUT_DIR/gpt5nano/gpt5nano_summary.json" | python -m json.tool
+echo ""
+echo "Qwen3-VL-8B (synth-ai):"
+cat "$OUTPUT_DIR/qwen3vl/qwen_vl_summary.json" | python -m json.tool
+echo ""
+echo "Frames saved in:"
+echo "  - $OUTPUT_DIR/gpt5nano/gpt5nano_frames/"
+echo "  - $OUTPUT_DIR/qwen3vl/qwen_vl_frames/"

examples/qwen_vl/run_vision_sft_pipeline.sh

@@ -0,0 +1,175 @@
+#!/bin/bash
+# Complete pipeline: Collect vision traces → Filter → Train SFT
+# Uses synth-ai CLI tools for data collection and processing
+
+set -e
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+# Allow callers to override root paths, otherwise derive them relative to this script.
+SYNTH_DIR="${SYNTH_DIR:-$(cd "$SCRIPT_DIR/../.." && pwd)}"
+DEFAULT_MONOREPO_DIR="$(cd "$SYNTH_DIR/.." && pwd)/monorepo"
+MONOREPO_DIR="${MONOREPO_DIR:-$DEFAULT_MONOREPO_DIR}"
+
+if [ ! -d "$SYNTH_DIR" ]; then
+    echo "Error: synth-ai repository not found at: $SYNTH_DIR"
+    exit 1
+fi
+
+if [ ! -d "$MONOREPO_DIR" ]; then
+    echo "Warning: MONOREPO_DIR not found at: $MONOREPO_DIR"
+    echo "         Set MONOREPO_DIR to a valid path if you plan to run the optional training step."
+fi
+
+# Configuration
+MODEL="gpt-5-nano"
+PROVIDER="openai"
+NUM_EPISODES=100
+OUTPUT_DIR="traces/gpt5nano_vision"
+
+echo "======================================"
+echo "Vision SFT Pipeline for Crafter"
+echo "======================================"
+echo ""
+echo "Model: $MODEL"
+echo "Provider: $PROVIDER"
+echo "Episodes: $NUM_EPISODES"
+echo "Output: $OUTPUT_DIR"
+echo ""
+
+# Check API keys
+if [ "$PROVIDER" = "openai" ]; then
+    if [ -z "$OPENAI_API_KEY" ]; then
+        echo "Error: OPENAI_API_KEY not set"
+        exit 1
+    fi
+    echo "✓ OpenAI API key found"
+elif [ "$PROVIDER" = "synth" ]; then
+    if [ -z "$SYNTH_API_KEY" ]; then
+        echo "Error: SYNTH_API_KEY not set"
+        exit 1
+    fi
+    echo "✓ Synth API key found"
+fi
+
+if [ -z "$BACKEND_BASE_URL" ]; then
+    echo "Warning: BACKEND_BASE_URL not set, using default"
+    export BACKEND_BASE_URL="https://synth-backend-dev-docker.onrender.com/api"
+fi
+
+echo ""
+
+# Step 1: Collect traces
+echo "======================================"
+echo "STEP 1: Collect Vision Traces"
+echo "======================================"
+echo ""
+echo "Running $NUM_EPISODES episodes with $MODEL..."
+echo "This will take ~30-60 minutes"
+echo ""
+
+cd "$SYNTH_DIR"
+
+uvx synth-ai eval \
+    --config examples/qwen_vl/configs/eval_${PROVIDER}_${MODEL/\//_}_vision.toml \
+    --output-dir "$OUTPUT_DIR" \
+    || {
+        # Fallback to gpt5nano config if custom config not found
+        uvx synth-ai eval \
+            --config examples/qwen_vl/configs/eval_gpt5nano_vision.toml \
+            --output-dir "$OUTPUT_DIR"
+    }
+
+echo ""
+echo "✅ Trace collection complete!"
+echo ""
+
+# Step 2: Filter and export to SFT format
+echo "======================================"
+echo "STEP 2: Filter & Export to SFT JSONL"
+echo "======================================"
+echo ""
+
+uvx synth-ai filter \
+    --config examples/qwen_vl/configs/filter_vision_sft.toml \
+    --input-db "$OUTPUT_DIR/rollouts.db" \
+    --output-dir "$OUTPUT_DIR/sft"
+
+echo ""
+echo "✅ Filtering complete!"
+echo ""
+
+# Show dataset stats
+echo "======================================"
+echo "Dataset Statistics"
+echo "======================================"
+echo ""
+
+if [ -f "$OUTPUT_DIR/sft/filter_stats.json" ]; then
+    cat "$OUTPUT_DIR/sft/filter_stats.json" | python3 -m json.tool
+else
+    echo "Train samples: $(wc -l < "$OUTPUT_DIR/sft/train.jsonl")"
+    echo "Val samples: $(wc -l < "$OUTPUT_DIR/sft/val.jsonl")"
+fi
+
+echo ""
+
+# Step 3: Train SFT (optional - user can run this separately)
+echo "======================================"
+echo "STEP 3: Train Vision SFT (Optional)"
+echo "======================================"
+echo ""
+echo "To train the model, run:"
+echo ""
+echo "  cd $MONOREPO_DIR"
+echo "  uvx synth-ai train \\"
+echo "    --type sft \\"
+echo "    --config configs/vision_sft/crafter_qwen3vl_8b_gpt5nano.toml \\"
+echo "    --dataset $SYNTH_DIR/$OUTPUT_DIR/sft/train.jsonl \\"
+echo "    --eval-dataset $SYNTH_DIR/$OUTPUT_DIR/sft/val.jsonl \\"
+echo "    --env-file backend/.env.dev"
+echo ""
+
+read -p "Run training now? (y/N) " -n 1 -r
+echo
+if [[ $REPLY =~ ^[Yy]$ ]]; then
+    echo ""
+    echo "Starting SFT training..."
+    echo ""
+    
+    if [ ! -d "$MONOREPO_DIR" ]; then
+        echo "Error: MONOREPO_DIR not found. Set MONOREPO_DIR to your monorepo path before running training."
+        exit 1
+    fi
+    
+    cd "$MONOREPO_DIR"
+    
+    uvx synth-ai train \
+        --type sft \
+        --config configs/vision_sft/crafter_qwen3vl_8b_gpt5nano.toml \
+        --dataset "$SYNTH_DIR/$OUTPUT_DIR/sft/train.jsonl" \
+        --eval-dataset "$SYNTH_DIR/$OUTPUT_DIR/sft/val.jsonl" \
+        --env-file backend/.env.dev
+    
+    echo ""
+    echo "✅ Training complete!"
+else
+    echo ""
+    echo "Skipping training. You can run it later using the command above."
+fi
+
+echo ""
+echo "======================================"
+echo "Pipeline Complete!"
+echo "======================================"
+echo ""
+echo "📂 Outputs:"
+echo "  - Raw traces: $OUTPUT_DIR/rollouts.db"
+echo "  - SFT train: $OUTPUT_DIR/sft/train.jsonl"
+echo "  - SFT val: $OUTPUT_DIR/sft/val.jsonl"
+echo "  - Stats: $OUTPUT_DIR/sft/filter_stats.json"
+echo ""
+echo "🚀 Next steps:"
+echo "  1. Train SFT model (see command above)"
+echo "  2. Evaluate trained model"
+echo "  3. Fine-tune with RL"
+echo ""