OpenAI API available in the opentrons-ai-server (#15169)

# Overview

This PR introduces a structured pattern for handling requests and
responses within the lambda. It demonstrates actual communication with
OpenAI, starting from the lambda and extending back to the client.
Additionally, the endpoint is secured, requiring a JWT bearer token for
access.
___

**Details:**
- **Client for Testing:** Established a dedicated live client to
facilitate testing.
- **Testing:** Developed tests that exercise the client and all
endpoints.


## Status

- [x] Deployed to dev and live tested `make build deploy test-live
ENV=dev AWS_PROFILE=the-profile`
- [x] Deployed to sandbox and live tested `make build deploy test-live
ENV=sandbox AWS_PROFILE=the-profile`
- [x] `make live-client` with fake=false works on sandbox
- [x] `make live client` with fake=false works on dev

> [!NOTE]
> to exercise the client reach out to @y3rsh for the `test.env` you need

opentrons-ai-server/.gitignore

@@ -3,3 +3,7 @@ results
 package
 function.zip
 requirements.txt
+test.env
+cached_token.txt
+tests/helpers/cached_token.txt
+tests/helpers/test.env

opentrons-ai-server/Makefile

@@ -21,15 +21,15 @@ black-check:
 
 .PHONY: ruff
 ruff:
-	python -m pipenv run python -m ruff check . --fix
+	python -m pipenv run python -m ruff check . --fix --unsafe-fixes
 
 .PHONY: ruff-check
 ruff-check:
 	python -m pipenv run python -m ruff check .
 
 .PHONY: mypy
 mypy:
-	python -m pipenv run python -m mypy aws_actions.py api
+	python -m pipenv run python -m mypy deploy.py api tests
 
 .PHONY: format-readme
 format-readme:
@@ -56,7 +56,7 @@ pre-commit: fixup unit-test
 
 .PHONY: gen-env
 gen-env:
-	python -m pipenv run python api/settings.py
+	python -m pipenv run python -m api.settings
 
 .PHONY: unit-test
 unit-test:
@@ -71,13 +71,13 @@ clean-package:
 .PHONY: gen-requirements
 gen-requirements:
 	@echo "Generating requirements.txt from Pipfile.lock..."
-	python -m pipenv requirements --hash > requirements.txt
+	python -m pipenv requirements > requirements.txt
 
 .PHONY: install-deps
 install-deps:
 	@echo "Installing dependencies to package/ directory..."
 	mkdir -p package
-	python -m pipenv run pip install -r requirements.txt --target ./package --upgrade
+	docker run --rm -v "$$PWD":/var/task "public.ecr.aws/sam/build-python3.12" /bin/sh -c "pip install -r requirements.txt -t ./package"
 
 .PHONY: package-lambda
 package-lambda:
@@ -96,9 +96,20 @@ ENV ?= sandbox
 .PHONY: deploy
 deploy:
 	@echo "Deploying to environment: $(ENV)"
-	python -m pipenv run python aws_actions.py --env $(ENV) --action deploy
+	python -m pipenv run python deploy.py --env $(ENV)
 
-.PHONY: test-lambda
-test-lambda:
-	@echo "Invoking the latest version of the lambda: $(ENV)"
-	python -m pipenv run python aws_actions.py --env $(ENV) --action test
+.PHONY: direct-chat-completion
+direct-chat-completion:
+	python -m pipenv run python -m api.domain.openai_predict
+
+.PHONY: print-client-settings-vars
+print-client-settings-vars:
+	python -m pipenv run python -m tests.helpers.settings
+
+.PHONY: live-client
+live-client:
+	python -m pipenv run python -m tests.helpers.client
+
+.PHONY: test-live
+test-live:
+	python -m pipenv run python -m pytest tests -m live --env $(ENV)

opentrons-ai-server/Pipfile

@@ -18,6 +18,8 @@ aws-lambda-powertools = {extras = ["all"], version = "==2.37.0"}
 boto3 = "==1.34.97"
 boto3-stubs = "==1.34.97"
 rich = "==13.7.1"
+pyjwt = "==2.8.0"
+cryptography = "==42.0.7"
 
 [requires]
 python_version = "3.12"

opentrons-ai-server/README.md

@@ -18,7 +18,7 @@ The Opentrons AI application's server.
    1. This allows formatting of of `.md` and `.json` files
 1. select the python version `pyenv local 3.12.3`
    1. This will create a `.python-version` file in this directory
-1. select the node version `nvs` currently 18.19\*
+1. select the node version with `nvs` or `nvm` currently 18.19\*
 1. Install pipenv and python dependencies `make setup`
 
 ## Install a dev dependency
@@ -31,11 +31,11 @@ The Opentrons AI application's server.
 
 ## Stack and structure
 
-### Lambda Pattern
+### Tools
 
 - [powertools](https://powertools.aws.dev/)
-- [reinvent talk for the pattern](https://www.youtube.com/watch?v=52W3Qyg242Y)
-- [for creating docs](https://www.ranthebuilder.cloud/post/serverless-open-api-documentation-with-aws-powertools)
+- [pytest]: https://docs.pytest.org/en/
+- [openai python api library]: https://pypi.org/project/openai/
 
 ### Lambda Code Organizations and Separation of Concerns
 
@@ -46,10 +46,13 @@ The Opentrons AI application's server.
 - integration
   - the integration with other services
 
-[pytest]: https://docs.pytest.org/en/
-[openai python api library]: https://pypi.org/project/openai/
+## Dev process
 
-## Deploy
+1. Make your changes
+1. Fix what can be automatically then lent and unit test like CI will `make pre-commit`
+1. `make pre-commit` passes
+1. deploy to sandbox `make build deploy test-live ENV=sandbox AWS_PROFILE=the-profile`
 
-1. build the package `make build`
-1. deploy the package `make deploy ENV=sandbox AWS_PROFILE=robotics_ai_sandbox`
+## TODO
+
+- llama-index is gigantic. Have to figure out how to get it in the lambda

opentrons-ai-server/api/domain/openai_predict.py

@@ -0,0 +1,55 @@
+from typing import List
+
+from openai import OpenAI
+from openai.types.chat import ChatCompletion, ChatCompletionMessage, ChatCompletionMessageParam
+
+from api.domain.prompts import system_notes
+from api.settings import Settings, is_running_on_lambda
+
+
+class OpenAIPredict:
+    def __init__(self, settings: Settings) -> None:
+        self.settings: Settings = settings
+        self.client: OpenAI = OpenAI(api_key=settings.openai_api_key.get_secret_value())
+
+    def predict(self, prompt: str, chat_completion_message_params: List[ChatCompletionMessageParam] | None = None) -> None | str:
+        """The simplest chat completion from the OpenAI API"""
+        top_p = 0.0
+        messages: List[ChatCompletionMessageParam] = [{"role": "system", "content": system_notes}]
+        if chat_completion_message_params:
+            messages += chat_completion_message_params
+
+        user_message: ChatCompletionMessageParam = {"role": "user", "content": f"QUESTION/DESCRIPTION: \n{prompt}\n\n"}
+        messages.append(user_message)
+
+        response: ChatCompletion = self.client.chat.completions.create(
+            model=self.settings.OPENAI_MODEL_NAME,
+            messages=messages,
+            stream=False,
+            temperature=0.005,
+            max_tokens=4000,
+            top_p=top_p,
+            frequency_penalty=0,
+            presence_penalty=0,
+        )
+
+        assistant_message: ChatCompletionMessage = response.choices[0].message
+        return assistant_message.content
+
+
+def main() -> None:
+    """Intended for testing this class locally."""
+    if is_running_on_lambda():
+        return
+    from rich import print
+    from rich.prompt import Prompt
+
+    settings = Settings.build()
+    openai = OpenAIPredict(settings)
+    prompt = Prompt.ask("Type a prompt to send to the OpenAI API:")
+    completion = openai.predict(prompt)
+    print(completion)
+
+
+if __name__ == "__main__":
+    main()

opentrons-ai-server/api/domain/prompts.py

@@ -0,0 +1,37 @@
+system_notes = """\
+You are an expert at generating a protocol based on Opentrons Python API v2.
+You will be shown the user's question/description and information related to
+the Opentrons Python API v2 documentation. And you respond the user's question/description
+using only this information.
+
+INSTRUCTIONS:
+
+1) All types of protocols are based on apiLevel 2.15,
+ thus prepend the following code block
+`metadata` and `requirements`:
+```python
+from opentrons import protocol_api
+
+metadata = {
+    'protocolName': '[protocol name by user]',
+    'author': '[user name]',
+    'description': "[what is the protocol about]"
+}
+requirements = {"robotType": "[Robot type]", "apiLevel": "2.15"}
+```
+
+2) See the transfer rules <<COMMON RULES for TRANSFER>> below.
+3) Learn examples see <<EXAMPLES>>
+
+4) Inside `run`  function, according to the description generate the following in order:
+- modules
+- adapter
+- labware
+- pipettes
+Note that sometimes API names is very long eg.,
+`Opentrons 96 Flat Bottom Adapter with NEST 96 Well Plate 200 uL Flat`
+
+
+5)  If the pipette is multi-channel eg., P20 Multi-Channel Gen2, please use `columns` method.
+\n\n\
+"""