Product Vision
Suite of tools to facilitate enterprise execution
Ship solution that helps people focus on business value
- software
- hardware
Local-first (easily deployable & online-ready) suite of tools including frontends, backends, language models to facilitate enterprise execution with AI.
Build & deploy AI apps with a mobile phone.
- resonsive markup for all pages + good UX
Execute AI Enterprise with a mobile phone.
Enterprise Runtime Stack
- Enterprise-helper (entities)
- Continuous training pipeline
- Knowledge base
- Ticket system
- software engineer AI
- DevOps platform
- CI-server
- CD-server
- AI-server
- Code Intelligence
- Network intelligence
- Enterprise Intelligence
- Team Communication
- Status Page System
- etc.
LLM-empowered features
AI-native enterprise execution platform. Each feature, each process is to be thought of from the perspective of how it can be enhanced with LLMs. Throughout the entire Enterprise Runtime Stack.
Neural interface
Implement a neural interface to interact with LMs. Instead of sending text / voice to LMs, make LMs able to recognize thoughts and execute them. (e.g. neuralink ?)
bci (brain-computer-interface)
Responsible AI
SAFe has a bit analog approach to Responsible AI. My vision is just implement all evaluation types that are possible in programmatic way with LMs and AI Agents. So that any developer of an AI tool can:
- see the benchmark dashboard in real time with a report with all evals based on studying his data and code, processes with LMs and outlines of mistakes
- & recommendations of how to fix them
- & one-click implementation of fixes (like with code linting errors)
- & risks or law violations, etc.
Make entire SAFe & Responsible AI thing just a program, which is easy to run locally for your project. Functional enough to provide real value to developers of AI. Fast enough to be performant.
SRE ?
Consider adding SRE management to the Enterprise-helper?
- The Evolving SRE Engagement Model (opens in a new tab)
- Extract vision from ^^
Solutions / Features
-
The AI-Native Software Delivery Platform (opens in a new tab)
- Shaping the Future of AI-Native Software Delivery (opens in a new tab)
- AI DevOps assistant (opens in a new tab)
- study all of Harness coolness & get inspiration from
- fill in kb with ideas, tasks
- Harness named a Leader in the 2024 Gartner® Magic Quadrant™ for DevOps Platforms (opens in a new tab)
- Harness named a Leader in the 2024 Gartner® Magic Quadrant™ for DevOps Platforms (opens in a new tab)
- github.com/harness (opens in a new tab)
- drone (opens in a new tab) / drone.io (opens in a new tab)
-
- @nxlv/python
- bolt.new-like app UI with MUI (for opensource community)
- github template reposities (opens in a new tab) ?
-
Entities
-
Prompt templates (?) - Should I use them, to have some context when sending a prompt from web page to the model? (DSPy?). E.g. each entity page to have it's own prompt template, so that there is context when a user prompt for a given entity is sent?
-
Web-containerize Enterprise-helper
-
Add IAM & SSO (Single-Sign-On)
-
Add team communication with element.io
- element-hq/element-web (opens in a new tab) github
- Add realtime translations? (enable people speaking different languages to communicate in their own language, and get translations in real-time (text / voice?)) Talk to Anyone! - Better Translations with OpenAI's NEW Realtime API (opens in a new tab)
-
Build a CLI-tool for Enterprise-helper?
-
add codeQL (opens in a new tab) for code security
-
Add code-intelligence tool (for code search, fix, update) across codebases
- with sourcegraph (opens in a new tab) or something similar (& maintained as open-source)
- https://github.com/sourcegraph/sourcegraph-public-snapshot (opens in a new tab)
-
Implement an education-portal (to help learn AI-development)
- with noodle.run (opens in a new tab) / noodle github (opens in a new tab) or similar
-
vpn?
-
database package manager ?
- https://database.dev/ (opens in a new tab) - For Postgres Trusted Language Extensions
- pg_tle (opens in a new tab) - Framework for building trusted language extensions for PostgreSQL
- PostgreSQL provides an extension framework for adding more functionality to PostgreSQL without having to fork the codebase. This powerful mechanism lets developers build new functionality for PostgreSQL, such as new data types, the ability to communicate with other database systems, and more. It also lets developers consolidate code that is functionally related and apply a version to each change. This makes it easier to bundle and distribute software across many unique PostgreSQL databases.
- https://supabase.github.io/dbdev/ (opens in a new tab)
- https://github.com/supabase/dbdev (opens in a new tab)
-
Use visulima (opens in a new tab) for knowledge base?
- Update existing kb with card links (opens in a new tab)
- add comments
- add footer
- add reactions
-
add badges to git repo with shields (opens in a new tab)
- code-coverage
- stable release version
- package manager release
- status of third-party dependencies
- static code analysis grade
- SemVer version observance
- Python (JS?) package downloads
- Uptime Robot percentage
- create own badges (opens in a new tab) if needed
-
add AI powered search (?)
-
automate syncing between your computer and cloud drive?
-
add a bookreader (opens in a new tab) for on-site library?
- openlibrary-client (opens in a new tab) ?
- free-programming-books (opens in a new tab) Python Client Library for the Archive.org OpenLibrary API
-
consider Ghost (opens in a new tab) for something?
-
typebot (opens in a new tab) self-hosted chatbot builder?
-
markdown powered blog - beam (opens in a new tab)
-
manage database with UI ? rowy (opens in a new tab)
-
support maths with mathjax (opens in a new tab) / katex (opens in a new tab) / ckeditor5-math (opens in a new tab) ?
-
add rich text editor (to knowledge base?)
- Quill (opens in a new tab) / gh (opens in a new tab)
- ckeditor 5 (opens in a new tab) / gh (opens in a new tab)
- TinyMCE (opens in a new tab) / gh (opens in a new tab)
- Slate (opens in a new tab) / slate (opens in a new tab)
- Froala (opens in a new tab) / gh (opens in a new tab)
- Summernote (opens in a new tab) / gh (opens in a new tab)
- ProseMirror (opens in a new tab) / gh (opens in a new tab)
- Trumbowyg (opens in a new tab) / gh (opens in a new tab)
-
Share conversations (chat) history across different LM-chat apps. (e.g. open-webui - Enterprise -helper).
-
Implement AI Assistance for Finance
- Reference: Report Cruncher as a reference for Fin part of Enterprise-helper: https://chuangtc.com/openai-hackathon-2023/ (opens in a new tab)
-
Add Enterprise-helper AI plugins for business use-cases:
- https://tome.app/ (opens in a new tab)
- https://zapier.com/ (opens in a new tab)
- https://gravitywrite.com/ (opens in a new tab)
- https://audiobox.metademolab.com/ (opens in a new tab)
- https://akool.com/ (opens in a new tab)
- https://elevenlabs.io/ (opens in a new tab)
- https://www.goenhance.ai/ (opens in a new tab)
- https://pictory.ai/ (opens in a new tab)
- https://taplio.com/ (opens in a new tab)
- https://www.nvidia.com/nl-nl/geforce/broadcasting/broadcast-app/ (opens in a new tab)
- https://revoicer.com/ (opens in a new tab)
-
Knowledge packages for LMs ?
-
Add LLM benchmark:
-
Add LLM Sentinel:
- https://github.com/aidatatools/LLM_Sentinel (opens in a new tab)
- Read Articles, Extract !!!!
-
ADD click-template-cli
-
Update product vision. What I'm building is not just an application with features, but rather intelligence, applicable to certain domains, e.g. code intelligence, finance intelligence, quality intelligence, etc. Thus, the Enterprise-helper is a set of programmatic intelligences, capable of human-like problem-solving.
-
Consider ghosty (opens in a new tab) as a terminal?
Finance
Enable AI-ified insights into finance & possible optimizations.
Language Models
- Integrated solution, that would be capable of
- Access and operations with the Operating System and File System
- CRUD of files
- Writing code
- Terminal access (executing commands)
- Browsing
- Running LMs locally
- Knowledge base with UI (Confluence analog)
- Ticket system with UI (JIRA analog)
- SAFe system with UI
- How do I make it?
- Access and operations with the Operating System and File System
- tools for different purposes (like Vscode for coding, or Figma for UI prototyping might be released at some point bundled with LMs, trained for performance in their corresponding domain). Most probably, it is a matter of time, when it becomes a new industry standard.
What does it mean for the product I'm building?
Users collaboration
- Users can
- work with the product fully locally (solo-mode)
- deploy the app to the server and work with the product together
- user access
How do I architect the data-persistence to satisfy both scenarios?
Architecture
- Data persistence
- Local Database to persist the data of the app.
- What else? Persistence of LM interactions?
- etc ?
Random
- Perplexity: Llama 3.1 Sonar 405B Online (opens in a new tab) $5/M input/output tokens
- Llama 3.2 90B Vision Instruct (opens in a new tab)
- Llama 3.2 11B Vision Instruct (free) (opens in a new tab) FREE!
- openrouter.ai/models (opens in a new tab)
- github.com/OpenRouterTeam (opens in a new tab)
- litellm openrouter (opens in a new tab)
Legal
- Build own LM from sratch or use an existing one and build on top of it?
- Can probably use Llama, based on Legal Considerations until the cutoff of 700M MAU. For fastest TTM.
- Consider building own (opens in a new tab) when approaching this point? Or from the very beginning? For full control & flexibility + legal rights (intellectual property) ?
POCs
List, what are current POCs, what I intent to achieve with them.
- Dev POCs
- Dev POC #1
- Strategic Themes + Chat with docs (?)
- Dev POC #2
- Dev POC #1 + RAG + WebSearch (?)
- Dev POC #3.1
- Put together
- knowledge base
- ticket system
- software-engineer AI
- continuous training pipeline
- DevOps platform
- Status page system
- Put together
- Dev POC #3.2
- Implement a chain of entities
- Strategic themes
- Epics
- Capabilities
- Features
- User Stories
- Implement a chain of entities
- Dev POC 4
- continuous training pipeline (litgpt)
- Dev POC 5
- chat + web-container (dev env: code IDE, browser, terminal)
- Dev POC 6
- generate app from strategic themes
- repo-context.json + LM + Strategic Themes
- generate app from strategic themes
- Dev POC #1
- Training POCs
- Training POC #1
- Fine-tuned Llama 3.1 8B
- Training POC #1
LM Evaluation for SAFe
- Create own evaluation for LMs for SAFe questions.
- Create a dataset of Q&A.
- Base
- Supervised
- Create a dataset of Q&A.
- Evaluate models against it.
- Get the results.
- Compare
- Which LM performs best on this benchmark (give it a name) ?