Manual GPT (remote) instrumentation example

examples/chatgpt_example.py

@@ -0,0 +1,113 @@
+# /// script
+# requires-python = ">=3.10"
+# dependencies = ["wildedge-sdk", "openai"]
+#
+# [tool.uv.sources]
+# wildedge-sdk = { path = "..", editable = true }
+# ///
+"""ChatGPT (OpenAI API) — fully manual integration.
+
+Shows how to instrument a remote LLM with no local model file.
+Tracks input/output token counts, generation config, latency, errors,
+and user feedback without any auto-instrumentation hooks.
+
+Run with: uv run chatgpt_example.py
+Requires: WILDEDGE_DSN and OPENAI_API_KEY environment variables.
+"""
+
+from openai import OpenAI
+
+import wildedge
+from wildedge import FeedbackType, GenerationConfig, GenerationOutputMeta, TextInputMeta
+from wildedge.timing import Timer
+
+MODEL = "gpt-4o"
+MODEL_VERSION = "2024-08-06"
+
+client = wildedge.WildEdge(
+    app_version="1.0.0",  # set WILDEDGE_DSN env var
+)
+
+# Remote models have no local object to inspect, so register with a
+# placeholder and supply all metadata explicitly.
+handle = client.register_model(
+    object(),
+    model_id=f"openai/{MODEL}",
+    source="https://api.openai.com",
+    family="gpt-4o",
+    version=MODEL_VERSION,
+)
+
+openai_client = OpenAI()  # set OPENAI_API_KEY env var or pass api_key= explicitly
+
+prompts = [
+    "Explain transformer attention in one sentence.",
+    "What is the capital of Japan?",
+    "Write a haiku about edge AI.",
+]
+
+temperature = 0.7
+max_tokens = 256
+
+for turn_index, prompt in enumerate(prompts):
+    messages = [{"role": "user", "content": prompt}]
+
+    try:
+        with Timer() as t:
+            response = openai_client.chat.completions.create(
+                model=MODEL,
+                messages=messages,
+                temperature=temperature,
+                max_tokens=max_tokens,
+            )
+
+        usage = response.usage
+        choice = response.choices[0]
+        completion = choice.message.content or ""
+        tokens_per_second = (
+            round(usage.completion_tokens / t.elapsed_ms * 1000, 1)
+            if usage.completion_tokens and t.elapsed_ms > 0
+            else None
+        )
+
+        inference_id = handle.track_inference(
+            duration_ms=t.elapsed_ms,
+            input_modality="text",
+            output_modality="text",
+            success=True,
+            input_meta=TextInputMeta(
+                char_count=len(prompt),
+                word_count=len(prompt.split()),
+                token_count=usage.prompt_tokens,
+                prompt_type="chat",
+                turn_index=turn_index,
+                contains_code="```" in prompt,
+            ),
+            output_meta=GenerationOutputMeta(
+                tokens_in=usage.prompt_tokens,
+                tokens_out=usage.completion_tokens,
+                tokens_per_second=tokens_per_second,
+                stop_reason=choice.finish_reason,
+                context_used=usage.total_tokens,
+            ),
+            generation_config=GenerationConfig(
+                temperature=temperature,
+                max_tokens=max_tokens,
+            ),
+        )
+
+        print(f"Q: {prompt}\nA: {completion}\n")
+
+        # Simulate feedback: short completions get a thumbs down.
+        feedback_type = (
+            FeedbackType.THUMBS_UP
+            if len(completion) > 40
+            else FeedbackType.THUMBS_DOWN
+        )
+        handle.track_feedback(inference_id, feedback_type)
+
+    except Exception as exc:
+        handle.track_error(error_code="UNKNOWN", error_message=str(exc)[:200])
+        raise
+
+client.close(timeout=5.0)