Merge pull request #266 from FullStackWithLawrence/next

incremental improvements to openai function calling for personal info

Author: lpm0073

README.md

@@ -17,33 +17,39 @@
 
 A [React](https://react.dev/) + [AWS Serverless](https://aws.amazon.com/serverless/) full stack implementation of the [30 example applications](https://platform.openai.com/examples) found in the official OpenAI API documentation. This repository is used as an instructional tool for the YouTube channel "[Full Stack With Lawrence](https://youtube.com/@FullStackWithLawrence)" as well as for University of British Columbia course, "[Artificial Intelligence Cloud Technology Implementation](https://extendedlearning.ubc.ca/courses/artificial-intelligence-cloud-technology-implementation/mg202)" taught by Lawrence McDaniel.
 
+_New in version 0.10.x: A new chat app named "OpenAI Function Calling". See [lambda_openai_function](https://github.com/FullStackWithLawrence/aws-openai/tree/main/api/terraform/python/openai_api/lambda_openai_function) for examples including the fully implemented "[get_current_weather()](https://platform.openai.com/docs/guides/function-calling)" from The official OpenAI API documentation, and also a fun example of how get OpenAI to not only recognize you but also say flowery nice things about you!_
+
 ![Marv](https://cdn.lawrencemcdaniel.com/marv.gif)
 
-**IMPORTANT DISCLAIMER: AWS' Lambda service has a hard 29-second timeout. OpenAI API calls often take longer than this, in which case the AWS API Gateway endpoint will return a 504 "Gateway timeout error" response to the React client. This happens frequently with apps created using chatgpt-4. Each of the 30 OpenAI API example applications are nonetheless implemented exactly as they are specified in the official documentation.**
+IMPORTANT DISCLAIMERS:
+
+1. AWS' Lambda service has a hard 29-second timeout. OpenAI API calls often take longer than this, in which case the AWS API Gateway endpoint will return a 504 "Gateway timeout error" response to the React client. This happens frequently with apps created using chatgpt-4. Each of the 30 OpenAI API example applications are nonetheless implemented exactly as they are specified in the official documentation.
+
+2. Distribution upload packages for AWS Lambda functions as well as AWS Lambda Layers are limited to 50mb (and 250mb unzipped). Often, this poses serious limitations for Layers, which are intended to store your PyPi / NPM package dependencies. Note that incidentally, these code samples are also pretty code scaffolding for alternative Docker-based deployment strategies using Elastic Container Service and/or Elastic Kubernetes Service.
 
 Code composition as of Jan-2024:
 
 ```console
 -------------------------------------------------------------------------------
 Language                     files          blank        comment           code
 -------------------------------------------------------------------------------
-Markdown                        49            738              6           2257
-HCL                             24            308            601           1990
-Python                          17            420            437           1580
-YAML                            19            111            101           1089
-JavaScript                      39            111            126           1086
-JSX                              6             42             44            793
-CSS                              5             31             14            172
-make                             1             26             29            109
+HCL                             29            346            714           2324
+Markdown                        51            751              6           2277
+Python                          22            514            557           1955
+YAML                            20            114            109           1204
+JavaScript                      39            114            127           1085
+JSX                              6             45             47            856
+CSS                              5             32             14            180
+Text                             6             13              0            116
+make                             1             26             30            114
 INI                              2             15              0             70
 HTML                             2              1              0             65
-Text                             3              5              0             64
 Jupyter Notebook                 1              0            186             48
-Bourne Shell                     4             13             42             36
+Bourne Shell                     5             17             55             47
 TOML                             1              1              0             23
 Dockerfile                       1              4              4              5
 -------------------------------------------------------------------------------
-SUM:                           174           1826           1590           9387
+SUM:                           191          1,993          1,849         10,369
 -------------------------------------------------------------------------------
 ```
 
@@ -96,6 +102,11 @@ A REST API implementing each of the [30 example applications](https://platform.o
 - [NodeJS](https://nodejs.org/en/download): used with NPM for local ReactJS developer environment, and for configuring/testing Semantic Release.
 - [Docker Compose](https://docs.docker.com/compose/install/): used by an automated Terraform process to create the AWS Lambda Layer for OpenAI and LangChain.
 
+Optional requirements:
+
+- [Google Maps API key](https://developers.google.com/maps/documentation/geocoding/overview). This is used the OpenAI API Function Calling coding example, "[get_current_weather()](https://platform.openai.com/docs/guides/function-calling)".
+- [Pinecone API key](https://docs.pinecone.io/docs/quickstart). This is used for OpenAI API Embedding examples.
+
 ## Documentation
 
 Detailed documentation for each endpoint is available here: [Documentation](./doc/examples/)

api/postman/OpenAI.postman_collection.json

@@ -1422,17 +1422,17 @@
         ],
         "body": {
           "mode": "raw",
-          "raw": "{\n    \"input_text\": \"Marv, I'd like to introduce you to all the nice YouTube viewers.\",\n    \"chat_history\": []\n}",
+          "raw": "{\n    \"input_text\": \"XMarv, I'd like to introduce you to all the nice YouTube viewers.\",\n    \"chat_history\": []\n}",
           "options": {
             "raw": {
               "language": "json"
             }
           }
         },
         "url": {
-          "raw": "{{base_url}}/examples/default-marv-sarcastic-chat",
+          "raw": "{{base_url}}/examples/openai-function-calling",
           "host": ["{{base_url}}"],
-          "path": ["examples", "default-marv-sarcastic-chat"]
+          "path": ["examples", "openai-function-calling"]
         }
       },
       "response": [

api/terraform/python/openai_api/lambda_openai_function/README.md

@@ -0,0 +1,9 @@
+# OpenAI Function Calling examples
+
+## function_weather.py
+
+Fully implements the "[get_current_weather()](https://platform.openai.com/docs/guides/function-calling)" from The official OpenAI API documentation. OpenAI's documentation provides scaffolding for this feature, but falls short of actually providing code that retrieves location-based current weather forecasts.
+
+## function_refers_to.py
+
+This module demonstrates an alternative implementation of prompt behavior modification involving both Function Calling, plus, dynamic modifications to the system prompt. This example relies on [lambda_config.yaml](./lambda_config.yaml) for personalization data.

api/terraform/python/openai_api/lambda_openai_function/function_refers_to.py

@@ -5,6 +5,7 @@
 """
 import json
 
+import requests
 import yaml
 from openai_api.common.const import PYTHON_ROOT
 from openai_api.lambda_openai_function.natural_language_processing import does_refer_to
@@ -13,6 +14,13 @@
 with open(PYTHON_ROOT + "/openai_api/lambda_openai_function/lambda_config.yaml", "r", encoding="utf-8") as file:
     lambda_config = yaml.safe_load(file)
 
+client_list_url = "https://api.lawrencemcdaniel.com/wp-json/wp/v2/posts?categories=46&tags=55&_embed&per_page=100"
+try:
+    client_list_response = requests.get(client_list_url, timeout=60)
+# pylint: disable=broad-except
+except Exception as e:
+    client_list_response = None
+
 
 def search_terms_are_in_messages(messages: list, search_terms: list = None, search_pairs: list = None) -> bool:
     """
@@ -52,15 +60,46 @@ def customized_prompt(messages: list) -> list:
     return messages
 
 
-def get_additional_info(inquiry_type: str) -> str:
-    """Return a table of URLs and other info about the location"""
+def get_client_list() -> list:
+    """Return a list of clients"""
+    if not client_list_response:
+        return []
+
+    if client_list_response.status_code == 200:
+        client_list = client_list_response.json()
+        client_list = [
+            client["title"]["rendered"] for client in client_list if "title" in client and "rendered" in client["title"]
+        ]
+        return client_list
+    return []
+
+
+def get_additional_info(inquiry_type: str = "biographical_info") -> str:
+    """Return select info from lambda_config.yaml"""
+    inquiry_type = inquiry_type or "biographical_info"
+    lambda_config["clients"] = get_client_list()
+
+    if inquiry_type == "client_list":
+        return json.dumps(lambda_config["clients"])
+
+    if inquiry_type == "contact_info":
+        return json.dumps(lambda_config["contact_information"])
+
+    if inquiry_type == "educational_info":
+        return json.dumps(
+            {
+                "education": lambda_config["profile"]["education"],
+                "certifications": lambda_config["profile"]["certifications"],
+            }
+        )
+
     if inquiry_type == "biographical_info":
-        return lambda_config["biographical_info"]
+        return json.dumps(lambda_config)
 
     if inquiry_type == "marketing_info":
         return json.dumps(lambda_config["marketing_info"])
 
-    return json.dumps({"error": "inquiry_type not recognized"})
+    return json.dumps(lambda_config)
 
 
 def info_tool_factory():
@@ -72,11 +111,20 @@ def info_tool_factory():
             "type": "function",
             "function": {
                 "name": "get_additional_info",
-                "description": "Get additional information about Lawrence McDaniel, full stack web developer and host of YouTube channel FullStackwithLawrence.",
+                "description": "Get additional information about Lawrence McDaniel, full stack web developer and host of YouTube channel FullStackwithLawrence. returns a personal bio, contact information, marketing information, client list, education background, professional certifications, etc.",
                 "parameters": {
                     "type": "object",
                     "properties": {
-                        "inquiry_type": {"type": "string", "enum": ["biographical_info", "marketing_info"]},
+                        "inquiry_type": {
+                            "type": "string",
+                            "enum": [
+                                "biographical_info",
+                                "marketing_info",
+                                "contact_info",
+                                "educational_info",
+                                "client_list",
+                            ],
+                        },
                     },
                     "required": ["inquiry_type"],
                 },

api/terraform/python/openai_api/lambda_openai_function/lambda_config.yaml

@@ -80,11 +80,19 @@ system_prompt: >
   Lawrence's Twitter is @FullStackWLarry. His email is [email protected].
   Lawrence's LinkedIn is https://www.linkedin.com/in/lawrencemcdaniel/.
   Lawrence's instructor profile at University of British Columbia is https://extendedlearning.ubc.ca/about-us/our-instructors/lawrence-mcdaniel
-
+contact_information:
+  - name: Lawrence P. McDaniel
+  - title: Full Stack Developer
+  - location: Mexico City
+  - website: https://lawrencemcdaniel.com/contact
+  - payments: https://clients.lawrencemcdaniel.com
+  - linkedin: https://www.linkedin.com/in/lawrencemcdaniel/
+  - whatsapp: +16178346172
+  - email: [email protected]
 marketing_info:
   name: Lawrence McDaniel
   title: Full Stack Developer
-  location: Cambridge, MA
+  location: Mexico City
   website: https://lawrencemcdaniel.com
   blog: https://blog.lawrencemcdaniel.com
   payments: https://clients.lawrencemcdaniel.com
@@ -93,10 +101,76 @@ marketing_info:
   twitter: https://twitter.com/FullStackWLarry
   youtube: https://www.youtube.com/@FullStackWithLawrence
   bio: https://extendedlearning.ubc.ca/about-us/our-instructors/lawrence-mcdaniel
-
+  flickr: https://www.flickr.com/photos/159844232@N02/
 biographical_info: >
   Lawrence McDaniel is a freelance full-stack web developer and online instructor at University of British Columbia. He has a BS in computer science and mathematics with minors in physics and English from University of North Texas. He was an early employee at three startups that were later acquired by Goldman Sachs, IBM and Deutsche Bank respectively, and has participated in a successful NASDAQ IPO. He has worked as a freelance technology consultant since 1999 and has advised dozens of startups as well as banks, hedge funds and real estate investment funds from around the world including Morgan Stanley, Fortress Investment Group, IBM Seterus, and others.
   He has expertise in data science, financial modeling and forecasting systems, analytics, machine learning, AI, big data, web, mobile, database technology, and Amazon Web Services (AWS) cloud infrastructure. He has 20+ years experience designing high performance, always-up transactional database systems using MySQL, MSSQL Server, Aurora and MongoDB. He also has years of expertise designing high availability, horizontally scalable cloud-based infrastructure environments. In fact, Lawrence is an ambassador for the AWS Activate Program in Mexico, helping startups and early-stage ventures get digital products and services to market using AWS. He advises companies on migration plans to AWS as well as works with early-stage ventures creating new highly scalable back-end environments and implementing continuous integration strategies.
   Lawrence is an angel investor, entrepreneur and company mentor with nearly 25 years of international experience with an emphasis on technology and US-Mexico trade. He has extensive startup, fund raising, business development and M&A experience with alternative investment and disruptive FinTech, edTech and IoT ventures. He is a mentor at Startup Mexico and Angel Ventures in Mexico City, and sits on the boards of edMex and M-Arca Foundation.
   But he’s no stranger to real work either! his family built the house where he grew up in rural northeast Texas, raised their own livestock, and farmed their own crops. He paid his way through college by working for six years during and after high school in various jobs as a printer, cattle hand, carpenter, roofer, sheet rocker, painter, glazier, welder, lumber yard worker, forklift mechanic, truck loader, landscaper and gardener, hot tub installer, janitor, grocery shelf stocker, carpet & commercial floor cleaner, dishwasher, waiter, and cook. During college he worked as a cafeteria food server, a tutor in the university math lab, a calculus paper grader and substitute lecturer for the math department and as a lab assistant for the physics department. Immediately after graduating college he helped launch one the first indoor rock climbing gyms, Exposure Indoor Rock Climbing gym in Carrollton, Texas, while simultaneously covering progress of NAFTA negotiations for a Dallas-based family office.
   Lawrence is fully English/Spanish bilingual and available to travel throughout all of North America.
+profile:
+  education:
+    - degree: Bachelors of Science
+      majors:
+        - Computer Science
+        - Mathematics
+      minors:
+        - Physics
+        - English
+      university: University of North Texas
+      year: 1992
+  skills:
+    - Full Stack Web Development
+    - Python
+    - ReactJS
+    - OpenAI API
+    - AWS
+    - Kubernetes
+    - Docker
+    - MySQL
+    - MongoDB
+  certifications:
+    - title: Data Science & Machine Learning - Making Data-driven Decisions
+      issuer: Massachusetts Institute of Technology
+      year: 2022
+      link: https://verify.mygreatlearning.com/verify/OHMJTCTD
+    - title: "LFS158x: Introduction to Kubernetes"
+      issuer: Linux Foundation
+      year: 2022
+      link: https://courses.edx.org/certificates/79888712a08642a6b35c9c077d354bb6
+    - title: "Front-End Web Development with React by Hong Kong University of Science and Technology"
+      issuer: Coursera
+      year: 2020
+      link:
+    - title: Accounting Analytics
+      issuer: Wharton School of Business
+      year: 2017
+      link: https://www.coursera.org/account/accomplishments/certificate/EB4FBYKJ2956
+    - title: Customer Analytics
+      issuer: Wharton School of Business
+      year: 2017
+      link: https://www.coursera.org/account/accomplishments/certificate/4YXEDYZJXQUD
+    - title: Operations Analytics
+      issuer: Wharton School of Business
+      year: 2017
+      link: https://www.coursera.org/account/accomplishments/certificate/7NSB8MFBWCT8
+    - title: People Analytics
+      issuer: Wharton School of Business
+      year: 2017
+      link: https://www.coursera.org/account/accomplishments/certificate/KZN42KMZ4HHJ
+    - title: Machine Learning
+      issuer: Stanford University
+      year: 2016
+      link: https://www.coursera.org/account/accomplishments/certificate/2MVW9YRMFYC9
+  projects:
+    - title: Cookiecutter Open edX
+      description: A community supported organization focused on providing reliable CI automation tools for Open edX™️ software deployed to Kubernetes with Tutor.
+      links:
+        - https://github.com/cookiecutter-openedx/
+      year: 2022
+    - title: Full Stack With Lawrence
+      description: A YouTube channel dedicated to teaching full stack web development.
+      links:
+        - https://www.youtube.com/@FullStackWithLawrence
+        - https://github.com/FullStackWithLawrence/
+      year: 2021