Forem: WellDefined.AI

Tigs: Never Lose That AI Context Again – Store Your LLM Chats in Git

Jinglei Ren — Thu, 02 Oct 2025 17:07:31 +0000

Originally published at blog.welldefined.ai

The biggest bug in software engineering isn’t a crash – it’s forgetting why.

That “god-tier” prompt you tuned yesterday? Gone today.

That design decision from three months ago? Nobody remembers why.

So when someone asks you something like 'why is this function designed this way?', in most cases your answer would be like:

🤷 “Uh… I think the AI suggested it?”

Tigs is here to fix that very human bug.

🧐 What is Tigs and why would you want it?

Tigs = Talks in Git → Specs.

In plain English: it captures your human ↔ AI conversations (prompts + responses), versions them alongside your commits in Git, and turns them into specs (which describes your system in a both comprehensive and precise way).

Working with AI in software development is never a one-shot deal. You don’t just write a single prompt and magically get perfect code back — it’s usually a messy dance of trial and error. One day you’re deep in a planning session with an AI — maybe you’ve kicked off plan mode to hash out a tricky feature. After three rounds of back-and-forth you finally arrive at a design that feels solid… only to find that a few days later the entire context has evaporated.

The same thing happens when you brainstorm new features with AI: a flurry of ideas, half-baked specs, design alternatives that spark in the moment but evaporate as soon as the chat window closes. And for newcomers stepping into a project — whether it’s a company repo or an open source library — nothing is more valuable than seeing not just the code, but the conversations and reasoning behind it. This is where Tigs shines: it stores those human-AI dialogues in Git notes, and your prompts, your design debates, your hard-won insights all become part of the project’s permanent, traceable history. The simple story is, with the assistance of Tigs:

Every brainstorm, every experiment, every failed attempt with an LLM → saved.
Those “throwaway” chats → distilled into specs and docs.
Everything goes into Git, right next to your code, with full version history.

So now:

👉 You never lose that prompt optimization again.

👉 Every “historical reason” actually has a record.

👉 Code and design docs finally stay in sync.

For individual developers:
- Preserve the reasoning around your code, not just the code itself.
- When you revisit a side project months later, you can pick it right back up.
- Show off a portfolio that includes not just code, but your thinking process.
For teams / open source:
- Transparent decisions — no more “oral agreements lost in chat.”
- Faster onboarding — newcomers see how ideas evolved.
- Less wasted effort — higher ROI on AI-assisted development.

Once, everyone knew the golden rule: Talk is cheap, show me the code.

In the AI era, the new rule has changed to: Code is cheap, show me the talk.

📚 Quickstart

# Install
pip install tigs

# In your repo
cd /path/to/your/repo

# Launch the TUI to curate chats and link them to commits
tigs store

# Review what’s linked
tigs view

# Push notes upstream (no commit rewrite)
tigs push

Take the TUI of tigs store for example. It is a 3-panel layout interactive window: The left panel shows the code commits in your current git repository; the right panel gives you all the log files to choose from; and the middle panel lists all the chats in the selected log file. Check the bottom status bar to see what each key press does.

👉 Open-source on GitHub: welldefined-ai/tigs – ⭐ Star us!

🌍 Why now?

Don't miss your chats until you miss them.

AI is part of daily dev life: prompts + conversations are replacing traditional development process.
Code ≠ the whole story: more and more “decisions” are made in chats with AI, not in code.
Info is fragmented: chats and experiments are scattered across AI agents: Claude Code, Gemini CLI, Qwen Code, Codex CLI, etc.

Tigs pulls all of that into Git and treats it as seriously as code. Think of it as: a design management tool for the LLM era.

🛣️ Specs and Roadmap

Tigs is still at an early stage — we’ve just finished building its core basic functionalities, and there’s plenty of room to grow! Over the next couple of months, we’ll be moving fast: polishing the TUI for a smoother developer experience, rolling out new features, and — most importantly — adding the missing Specs module, an essential piece of the puzzle that will make Tigs feel whole. When it comes to Specs, we already have a clear vision: they should be automatically generated by AI through reading both commits and chats, and serve as precise, comprehensive descriptions of the system’s state up to the current commit. Step by step, Tigs will evolve into a more complete tool for capturing, versioning, and building on your conversations with AI.

DevLake as a DevOps Data Monitoring Tool

Shubham Gupta — Sun, 19 Feb 2023 07:47:52 +0000

No matter what business or career you've chosen, data visualization can help by delivering data in the most efficient way possible. As one of the essential steps in the business intelligence process, data visualization takes the raw data, models it, and delivers the data so that conclusions can be reached.

Understanding the Importance of DevOps Data Visualization

In today's automated environments, there are panels that reserves the information about how an individual component/process is behaving at any particular instant compared to the past when we used to have instant meters which gave us instant readings. These automated systems help us to collect the raw data from their databases using plugins and display that data in real-time. Data visualization tools like grafana make the quick understanding to the user and give insight into any particular state of what is happening inside the system/component.
DevOps is a set of practices that collaborate communication of software development and IT operations team. It aims to shorten the systems development life cycle and provide continuous delivery with high software quality. While automating the process of Software delivery there could be certain tools that can be examined to provide situational insights.

At every stage of the diagram, we can take advantage of the analytical opportunities of that phase to gather meaningful metrics. Here is a list of the different phases and the corresponding metrics that could be monitored throughout the lifecycle:

How DevLake provides the Monitoring Solution?

Devlake is an open-source dev data platform that ingests, analyzes, and visualizes the fragmented data from DevOps tools to distil insights for engineering productivity. DevLake is designed for developer teams looking to make better sense of their development process and to bring a more data-driven approach to their own practices.

In simpler language devlake is a tool that helps us to collect our DevOps data, through various plugins that devlake stores in its repository. The data is further analyzed and we produce amazing graphical insights that would help to boost productivity. Let's take a simple example: Merico-dev is an organization that wants to keep track of all the commits/issues/PRs happening over the repository, maintaining and analyzing data would be a tedious process, but where devlake comes in, it would collect the data from GitHub, store the data in MySQL and visualize on grafana dashboard. In this way, Merico-dev can compare this month's productivity to the last month and could reward the developer with the most commits/issues/PRs merged.

A typical DevLake plugin's dataflow is illustrated below:

The Raw layer stores the API responses from data sources (DevOps tools) in JSON.
The Tool layer extracts raw data from JSONs into a relational schema that's easier to consume for analytical tasks.
The Domain layer attempts to build a layer of abstraction on top of the Tool layer so that analytics logic can be re-used across different tools.

So, when a pipeline is successfully run you get the complete collection of data you check out the scope of data from here. Successful pipelines can be analyzed on Dashboard like Grafana by using MySQL queries.

How to analyze DORA Metrics using DevLake

DevOps Research and Assessment (DORA) team has identified four key metrics that indicate the performance of a software development team the main idea is to figure out whether they are "low performers" to "elite performers". The four key metrics used are Deployment frequency (DF), Lead time for changes (LT), Mean time to recovery (MTTR), and Change failure rate (CFR).

Deployment Frequency - The frequency of the software release of an organization to the production level.
Lead Time for Changes - The amount of time an organization takes a commit/PR to get into the production level.
Change Failure Rate - The percentage of deployments causing a failure in production.
Time to Restore Service - How long it takes an organization to recover from a failure in production.

One of the biggest challenges of gathering these DORA metrics is solved by DevLake, currently, DevLake has 25+ plugins which help to collect the DevOps data from every stage of the SDLC and store it in a containerized database. The complete process can be automated by using CronJob ( i.e. DevLake BluePrint feature ).

Resulted Dashboard of DevLake

Calculating the Metrics

In this section, we would discuss how to translate the resulted data into the system-level calculation. This original research done by the DORA team surveyed real people rather than gathering systems data and bucketed metrics into a performance level, as follows:

Conclusion :

DevLake aims to become an ideal tool when it comes to analyzing any metrics of DevOps, through its extended plugins and robust data collecting system. It provides flexibility over the standard query and the user can create its dashboard.

References :

How do you measure productivity?

Maxim Wheatley — Tue, 15 Dec 2020 17:58:49 +0000

Hey Dev community! Maxim here from the Merico team. We spend a ton of time thinking about engineering productivity:
How to measure it? How to define it? How to improve it?

All simple questions with tricky answers. I'd love to hear from you on what you have used in the past that has worked whether tools, metrics, or techniques!

Hope to hear from you, and look forward to the conversation!

Disclosure: On Thursday we are releasing a free version of our analytics product to give developers better insights and data into their own performance and productivity. We are on a mission to help people articulate their accomplishments to become better coders and achieve more in their careers!