refactor and sort of finish sig gen

README.md

@@ -53,3 +53,11 @@ We call out to a live python environment from node, using a library called
 node-python bindings can see your poetry environment.
 
 The `node-calls-python` setup currently relies on a hard-coded
+
+## Installing models
+
+To add models as python packages:
+
+- Copy the `.whl` file to `/models` at the repo root
+- Add the file to poetry (you should be able to do
+  `poetry add models/<my-model>.whl`)

pyproject.toml

@@ -11,6 +11,8 @@ package-mode = false
 python = "3.11.6"
 requests = "^2.31.0"
 fastapi = "^0.110.2"
+spacy = "^3.7.4"
+en-core-web-sm = {path = "models/en_core_web_sm-3.7.1-py3-none-any.whl"}
 
 
 [build-system]

services/signature_generator/signature_generator.py

@@ -1,20 +1,18 @@
 import logging
 
-# If the environment is configured by poetry, how do we manage this?
-from fastapi import APIRouter
+from src.models import SignatureGenerator, SignatureInput
+from src.prompts import generate_prompt
+from src.utils import (
+    extract_api_info,
+    get_model_endpoint,
+    parse_openapi_spec,
+    trim_signature,
+)
 
 # TODO the platform should deal with logging
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
-# from models import SignatureGenerator, SignatureInput
-# from prompts import generate_prompt
-# from utils import (
-#     extract_api_info,
-#     get_model_endpoint,
-#     parse_openapi_spec,
-#     trim_signature,
-# )
 
 #async def main(data: SignatureInput) -> dict:
 
@@ -24,29 +22,33 @@
 class Payload:
   def __init__(self, dict):
     self.model = dict['model']
+    self.open_api_spec = dict['open_api_spec']
+    self.instruction = dict['instruction']
 
-# So thats interesting - data comes in as a dict
-# can I make it an object?
-# I would need to parse it basically
 def main(dataDict) -> dict:
     data = Payload(dataDict)
+
     logger.info(f"Generating signature for model {data.model}")
-    # generator = SignatureGenerator(get_model_endpoint(data.model))
-    # logger.info("Parsing OpenAPI specification")
-    # parsed_spec = parse_openapi_spec(data.open_api_spec)
-    # logger.info("Extracting API information from parsed spec with provided instruction")
-    # api_info = extract_api_info(parsed_spec, data.instruction)
-    # prompt_template = "signature" if data.model == "gpt3_turbo" else "signature_text"
-    # logger.info(f"Generating {data.model} prompt for signature generation")
-    # prompt = generate_prompt(
-    #     prompt_template, spec=api_info, instruction=data.instruction
-    # )
-
-    # signature = generator.generate(prompt)[0]
-    # signature = trim_signature(signature)
-
-    # logger.info("Signature generation complete")
-    # return {"signature": signature}
+    generator = SignatureGenerator(get_model_endpoint(data.model))
+
+    logger.info("Parsing OpenAPI specification")
+    parsed_spec = parse_openapi_spec(data.open_api_spec)
+
+    logger.info("Extracting API information from parsed spec with provided instruction")
+    api_info = extract_api_info(parsed_spec, data.instruction)
+
+    prompt_template = "signature" if data.model == "gpt3_turbo" else "signature_text"
+    logger.info(f"Generating {data.model} prompt for signature generation")
+
+    prompt = generate_prompt(
+        prompt_template, spec=api_info, instruction=data.instruction
+    )
+
+    signature = generator.generate(prompt)[0]
+    signature = trim_signature(signature)
+
+    logger.info("Signature generation complete")
+    return {"signature": signature}
     return data
 
 

services/signature_generator/test/test.py

@@ -0,0 +1 @@
+# tests go here