Python package improvements

Added an endpoint for getting the actual stored responses, and used it to test and improve the python package.

client-libs/python/openpipe/test_client.py

@@ -1,55 +1,106 @@
+from functools import reduce
 from dotenv import load_dotenv
-from . import openai, configure_openpipe
 import os
 import pytest
+from . import openai, configure_openpipe, configured_client
+from .api_client.api.default import local_testing_only_get_latest_logged_call
+from .merge_openai_chunks import merge_openai_chunks
+import random
+import string
+
+
+def random_string(length):
+    letters = string.ascii_lowercase
+    return "".join(random.choice(letters) for i in range(length))
+
 
 load_dotenv()
 
 openai.api_key = os.getenv("OPENAI_API_KEY")
 
 configure_openpipe(
-    base_url="http://localhost:3000/api", api_key=os.getenv("OPENPIPE_API_KEY")
+    base_url="http://localhost:3000/api/v1", api_key=os.getenv("OPENPIPE_API_KEY")
 )
 
 
+def last_logged_call():
+    return local_testing_only_get_latest_logged_call.sync(client=configured_client)
+
+
 def test_sync():
     completion = openai.ChatCompletion.create(
         model="gpt-3.5-turbo",
-        messages=[{"role": "system", "content": "count to 10"}],
+        messages=[{"role": "system", "content": "count to 3"}],
     )
 
-    print(completion.choices[0].message.content)
+    last_logged = last_logged_call()
+    assert (
+        last_logged.model_response.resp_payload["choices"][0]["message"]["content"]
+        == completion.choices[0].message.content
+    )
+    assert (
+        last_logged.model_response.req_payload["messages"][0]["content"] == "count to 3"
+    )
+
+    assert completion.openpipe.cache_status == "SKIP"
 
 
 def test_streaming():
     completion = openai.ChatCompletion.create(
         model="gpt-3.5-turbo",
-        messages=[{"role": "system", "content": "count to 10"}],
+        messages=[{"role": "system", "content": "count to 4"}],
         stream=True,
     )
 
-    for chunk in completion:
-        print(chunk)
+    merged = reduce(merge_openai_chunks, completion, None)
+    last_logged = last_logged_call()
+    assert (
+        last_logged.model_response.resp_payload["choices"][0]["message"]["content"]
+        == merged["choices"][0]["message"]["content"]
+    )
 
 
 async def test_async():
-    acompletion = await openai.ChatCompletion.acreate(
+    completion = await openai.ChatCompletion.acreate(
         model="gpt-3.5-turbo",
         messages=[{"role": "user", "content": "count down from 5"}],
     )
+    last_logged = last_logged_call()
+    assert (
+        last_logged.model_response.resp_payload["choices"][0]["message"]["content"]
+        == completion.choices[0].message.content
+    )
+    assert (
+        last_logged.model_response.req_payload["messages"][0]["content"]
+        == "count down from 5"
+    )
 
-    print(acompletion.choices[0].message.content)
+    assert completion.openpipe.cache_status == "SKIP"
 
 
 async def test_async_streaming():
-    acompletion = await openai.ChatCompletion.acreate(
+    completion = await openai.ChatCompletion.acreate(
         model="gpt-3.5-turbo",
         messages=[{"role": "user", "content": "count down from 5"}],
         stream=True,
     )
 
-    async for chunk in acompletion:
-        print(chunk)
+    merged = None
+    async for chunk in completion:
+        assert chunk.openpipe.cache_status == "SKIP"
+        merged = merge_openai_chunks(merged, chunk)
+
+    last_logged = last_logged_call()
+
+    assert (
+        last_logged.model_response.resp_payload["choices"][0]["message"]["content"]
+        == merged["choices"][0]["message"]["content"]
+    )
+    assert (
+        last_logged.model_response.req_payload["messages"][0]["content"]
+        == "count down from 5"
+    )
+    assert merged["openpipe"].cache_status == "SKIP"
 
 
 def test_sync_with_tags():
@@ -58,31 +109,54 @@ def test_sync_with_tags():
         messages=[{"role": "system", "content": "count to 10"}],
         openpipe={"tags": {"promptId": "testprompt"}},
     )
-    print("finished")
 
-    print(completion.choices[0].message.content)
+    last_logged = last_logged_call()
+    assert (
+        last_logged.model_response.resp_payload["choices"][0]["message"]["content"]
+        == completion.choices[0].message.content
+    )
+    print(last_logged.tags)
+    assert last_logged.tags["promptId"] == "testprompt"
+    assert last_logged.tags["$sdk"] == "python"
 
 
 def test_bad_call():
-    completion = openai.ChatCompletion.create(
-        model="gpt-3.5-turbo-blaster",
-        messages=[{"role": "system", "content": "count to 10"}],
-        stream=True,
+    try:
+        completion = openai.ChatCompletion.create(
+            model="gpt-3.5-turbo-blaster",
+            messages=[{"role": "system", "content": "count to 10"}],
+            stream=True,
+        )
+        assert False
+    except Exception as e:
+        pass
+    last_logged = last_logged_call()
+    print(last_logged)
+    assert (
+        last_logged.model_response.error_message
+        == "The model `gpt-3.5-turbo-blaster` does not exist"
     )
+    assert last_logged.model_response.status_code == 404
 
 
-@pytest.mark.focus
 async def test_caching():
+    messages = [{"role": "system", "content": f"repeat '{random_string(10)}'"}]
     completion = openai.ChatCompletion.create(
         model="gpt-3.5-turbo",
-        messages=[{"role": "system", "content": "count to 10"}],
+        messages=messages,
         openpipe={"cache": True},
     )
+    assert completion.openpipe.cache_status == "MISS"
+
+    first_logged = last_logged_call()
+    assert (
+        completion.choices[0].message.content
+        == first_logged.model_response.resp_payload["choices"][0]["message"]["content"]
+    )
 
     completion2 = await openai.ChatCompletion.acreate(
         model="gpt-3.5-turbo",
         messages=[{"role": "system", "content": "count to 10"}],
         openpipe={"cache": True},
     )
-
-    print(completion2)
+    assert completion2.openpipe.cache_status == "HIT"