Give an AI coding assistant persistent memory of failed approaches across separate chat sessions.
Store and recall engineering decisions and task handoff notes between chats.
Verify that an AI assistant correctly recalled and blocked a previously failed approach.
| anna-stolbovskaja/devsync | 0xallam/my-recipe | 0xhassaan/nn-from-scratch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | Python | Python | Python |
| Last pushed | — | 2022-11-22 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 2/5 | 2/5 | 4/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires a Groq API key and a Walrus Memory account to run the demo.
DevSync is a system prompt, meaning a set of instructions given to an AI assistant, designed to help that assistant remember engineering context across separate chat sessions. Normally, when a developer starts a new chat with an AI assistant, any lesson learned in a previous conversation is gone. DevSync tries to fix that by having the assistant store and later recall three specific kinds of information: approaches that were tried and failed, decisions that were explicitly settled on, and a short summary of where a task currently stands. For example, if a developer tries a certain technical approach and it fails for a specific reason, DevSync stores that failure and the reason behind it. Later, in a completely fresh chat, if the same approach comes up again, the assistant can recall that it was already tried and explicitly block it, rather than suggesting it again. Each of these three types of memory is kept in its own separate, clearly labeled storage area so they do not get mixed together, and the prompt is written to avoid ever storing passwords, private keys, source code, or personal information. The project includes a working demo written in Python that uses Groq's AI service to show the prompt in action across a two part conversation, first recalling nothing since it is a fresh session, then storing a failure, then showing that a second fresh session correctly recalls and blocks the same failed approach. A separate verification script checks that a completed demo run actually followed the intended rules correctly, matching stored records against what really happened. The repository also includes ten real memory records that were previously written using this system, each one viewable on a public blockchain based storage explorer, along with notes on which underlying AI models were tested and how they performed against the intended behavior. DevSync is written in Python, was built for a public prompt writing contest, and currently has 0 stars on GitHub. No license is stated in the material reviewed.
DevSync is a system prompt that gives AI coding assistants persistent memory of failures, decisions, and task handoffs across chats.
Mainly Python. The stack also includes Python, Groq, Walrus Memory.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.