improve readme and dockerfile (ngaut#63)

Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.10.12-bullseye
+FROM python:3.10.13
 
 WORKDIR /app
 
@@ -11,7 +11,8 @@ RUN pipenv install --deploy --ignore-pipfile --verbose
 # Install browsers
 RUN apt-get update && apt-get install -y \
     chromium-driver firefox-esr \
-    ca-certificates libnss3
+    ca-certificates libnss3 \
+    sqlite3 libsqlite3-dev
 
 COPY data/ ./data/
 COPY jarvis/ ./jarvis/

README.md

@@ -1,99 +1,131 @@
-# Jarvis: A Virtual Machine Tailored for AI
+# Jarvis: AI-Powered Virtual Machine
 
-Jarvis is a virtual machine designed to seamlessly integrate and run AI models. This README provides a step-by-step guide to setting up and using Jarvis.
+Welcome to Jarvis, a cutting-edge virtual machine designed specifically to facilitate AI tasks. This README offers a comprehensive guide to effectively set up and utilize Jarvis for optimal results.
+
+## Demo
+
+[Experimental Jarvis Application](/docs/media/demo.mp4)
 
 ## Prerequisites
 
-1. Ensure [pipenv](https://pypi.org/project/pipenv/) is installed on your system.
-2. Obtain the necessary API keys to interface with third-party services:
+1. Obtain the necessary API keys to interface with third-party services:
 - [OPENAI_API_KEY](https://platform.openai.com/account/api-keys)
-- Both `GOOGLE_API_KEY` and `GOOGLE_SEARCH_ENGINE_ID`` for integrating Google Search Engine.
+- Both `GOOGLE_API_KEY` and `GOOGLE_SEARCH_ENGINE_ID` for integrating Google Search Engine.
 
-## Setup
+## Installation and Setup
 
 1. Clone this repository:
+
 ```
 git clone https://github.com/ngaut/jarvis.git && cd jarvis
 ```
 
-2. Initiate a Python virtual environment:
+2. Set up the environment variables:
+
+- Rename `.env.template` to `.env`, and input the necessary API keys mentioned in the Prerequisites section.
+
+3. Build a Local Docker Image
+
 ```
-pipenv shell
+docker build -t jarvis-server:latest .
 ```
 
-3. Install the dependencies:
+4. Running Jarvis Server
+
+The Jarvis server operates on port `51155` by default, offering services via the gRPC protocol.
+
+To run the Jarvis server with Docker:
+
 ```
-pipenv install
+docker run --rm -p 51155:51155 \
+-v $(pwd)/workspace:/app/workspace \
+-v $(pwd)/.env:/app/.env \
+jarvis-server:latest
 ```
 
-4. Set up the environment variables:
-- Rename `.env.template` to `.env`, and input the necessary API keys mentioned in the Prerequisites section.
-- Update the `PYTHONPATH` environment variable with the absolute path of this project and include it in your .env file.
-
+Note: Ensure you've configured the `.env` file in the current directory before proceeding.
 
 ## Usage
 
-To generate a plan for tasks execution based on input goals:
+For guidance, you can refer to the code provided in this [demo](example.ipynb)
 
-```
-python -m jarvis --replan
-```
+1. Develop a Skill:
 
-To generate a plan for tasks execution based on a given goal file, which should be placed in the 'workspace' directory:
+Develop a skill that generates summaries of top stories from Hacknews.
 
-```
-python -m jarvis --replan --goalfile=<GOALFILE>
+```python
+stub.Execute(
+    jarvis_pb2.ExecuteRequest(
+        task=(
+            "Collect the top three articles featured on Hacker News (https://news.ycombinator.com/), "
+            "and produce a single professional reading summary that encompasses the content of all three articles, formatted in a user-friendly manner."
+        )
+    )
+)
 ```
 
-To translate all tasks in plan into JVM instructions:
+Task output example:
 
 ```
-python -m jarvis
+executor_id: "ea8fcfdf59c011002875a88fcdac5e97"
+task_id: 1
+task: Collect the top three articles featured on Hacker News (https://news.ycombinator.com/), and produce a single professional reading summary that encompasses the content of all three articles, formatted in a user-friendly manner.
+result: "The University of Turku in Finland is developing an artificial language corpus proficient in all European languages ..."
 ```
 
-To translate a task with specified task number into JVM instructions:
+2. Save a Skill:
 
-```
-python -m jarvis --compile=<task_num>
+A step-by-step guide to save a developed skill for subsequent use.
+
+```python
+stub.SaveSkill(
+    jarvis_pb2.SaveSkillRequest(
+        executor_id="ea8fcfdf59c011002875a88fcdac5e97",
+        skill_name="HackerNews top three articles summary",
+    )
+)
 ```
 
-To execute JVM instructions in the specified task:
+Task output example:
 
 ```
-python -m jarvis --yaml=1.yaml
-python -m jarvis --yaml=2.yaml
+executor_id: "ea8fcfdf59c011002875a88fcdac5e97"
+result: "skill is saved as HackerNews top three articles summary"
 ```
 
-## Docker Integration
 
-Run Jarvis within a Docker container by following these steps:
+3. Reuse Skills:
 
-### Build a Local Docker Image
+Recall and utilize previously saved skills for the same or related tasks.
 
-```
-docker build -t jarvis-server:latest .
+```python
+python run_skill_chain.py --workspace=workspace --skill_dir=skill_library --execution_dir=summary_hn_news --skills="HackerNews top three articles summary"
 ```
 
-Or pull a pre-built Docker image
+Task output example:
 
 ```
-docker pull public.ecr.aws/l7n5d1m0/jarvis-server:latest
-```
+executing skill: HackerNews top three articles summary
+--------------------------------------------------
+Skill Execution Summary
+--------------------------------------------------
 
-### Running Jarvis Server
+Skill Result: The article discusses a 3 state, 3 symbol Turing Machine called 'Bigfoot' that cannot be proven to halt or not without solving a Collatz-like problem ...
+Skill Error:  None
 
-The Jarvis server operates on port `51155` by default, offering services via the gRPC protocol.
+==================================================
+Detailed Task Infos
+==================================================
 
-To run the Jarvis server with Docker:
+Subtask: Collect the top three articles featured on Hacker News (https://news.ycombinator.com/), and produce a single professional reading summary that encompasses the content of all three articles, formatted in a user-friendly manner.
+Result: ...
+Error:   None
 
-```
-docker run --rm -p 51155:51155 \
--v $(pwd)/workspace:/app/workspace \
--v $(pwd)/.env:/app/.env \
-jarvis-server:latest
-```
+--------------------------------------------------
 
-Note: Ensure you've configured the `.env` file in the current directory before proceeding.
+End of Execution Summary
+--------------------------------------------------
+```
 
 ## Working with SuperAGI
 

example.ipynb

@@ -0,0 +1,70 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import grpc\n",
+    "import jarvis.server.jarvis_pb2 as jarvis_pb2\n",
+    "import jarvis.server.jarvis_pb2_grpc as jarvis_pb2_grpc\n",
+    "\n",
+    "channel = grpc.insecure_channel(\"localhost:51155\")\n",
+    "stub = jarvis_pb2_grpc.JarvisStub(channel)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "response=stub.Execute(\n",
+    "    jarvis_pb2.ExecuteRequest(\n",
+    "        task=(\n",
+    "            \"Collect the top three articles featured on Hacker News (https://news.ycombinator.com/), \"\n",
+    "            \"and produce a single professional reading summary that encompasses the content of all three articles, formatted in a user-friendly manner.\"\n",
+    "        )\n",
+    "    )\n",
+    ")\n",
+    "response"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "stub.SaveSkill(\n",
+    "    jarvis_pb2.SaveSkillRequest(\n",
+    "        executor_id=response.executor_id,\n",
+    "        skill_name=\"<customize skill name>\",\n",
+    "    )\n",
+    ")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "jarvis",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.11"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

jarvis/README.md

@@ -0,0 +1,42 @@
+# Overview
+
+Jarvis streamlines the process of task execution planning based on user-defined goals. Beyond that, it efficiently translates these tasks into JVM instructions, allowing seamless integration and execution.
+
+## How to Use Jarvis 
+
+### Generating a Task Execution Plan
+
+For Direct Input Goals. Generate a plan for task execution directly from provided input goals:
+
+```
+python -m jarvis --replan
+```
+
+Using a Goal File. If you have predefined goals stored in a file, place this file in the 'workspace' directory. To generate a plan based on this file:
+
+```
+python -m jarvis --replan --goalfile=<GOALFILE>
+```
+
+### Translating Tasks to JVM Instructions
+
+Translate All Tasks. Convert all tasks in the execution plan into their corresponding JVM instructions:
+
+```
+python -m jarvis
+```
+
+Translate Specific Task. If you wish to only translate a task with a particular task number:
+
+```
+python -m jarvis --compile=<task_num>
+```
+
+## Executing JVM Instructions
+
+Execute JVM instructions based on the specified task. Here are a couple of examples:
+
+```
+python -m jarvis --yaml=1.yaml
+python -m jarvis --yaml=2.yaml
+```

jarvis/smartgpt/instruction.py

@@ -140,7 +140,7 @@ def loop(self, jvm_instruction: JVMInstruction):
         # Extract the count and the list of instructions for the loop
         loop_count = 0
         # if loop_count is integer
-        print(
+        logging.info(
             f"loop instruction (seq={jvm_instruction.instruction.get('seq', 'N/A')}) args: {args}"
         )
         if isinstance(args["count"], int):