Repowise Foresight · Hosted and enterprise
Know what a change will need before you open a file.
History becomes prediction. Foresight is a set of models we trained on years of real commits and fixes across many open-source projects. Paste a bug report and it ranks the files the fix will touch. Open a change and it names what the change forgot. Every answer says why, in milliseconds, with no LLM.
Issue #4127 · open
Checkout sometimes charges twice after a timeout
On a slow connection the payment page times out, the customer tries again, and the card is charged twice.
payments/retry.pyfix lands here
41def charge_with_retry(order):42 try:43 return charge(order, key=new_key())44 except Timeout:45 return charge(order, key=new_key())The retry asks for a new idempotency key, so the second attempt becomes a second charge.
Keyword searchForesight · files the fix will touch
search rank
- 91
payments/retry.pyfixed for a similar timeout bug 3 weeks ago.71#9 - 72
billing/ledger.pychanges with retry.py 9 of 11 times.14#7 - 63
payments/gateway.pyraises the Timeout that retry.py catches.06#6 - 14
web/checkout_page.tsxmatches “checkout” 14 timesUI only, no payment calls.02#1 - 25
tests/test_checkout_ui.pymatches “checkout” 9 timestests the page, not the charge.01#2 - 36
README.mdmatches “charges” 3 times.01#3 - 47
web/checkout.cssmatches “checkout” 6 times.01#4 - 58
docs/checkout.mdmatches “timeout” 2 times.01#5
- No LLM
- 0.3 s on a laptop
- Explains every prediction
What it answers
The questions every change asks, answered first.
The open-source index knows what your code is. Foresight adds what a piece of work will need, ranked, with the reason next to each answer.
01
Where the fix goes
Paste a bug report or a stack trace. Foresight ranks the files, and the functions inside them, that the fix will touch.
e.g. payments/retry.py · fixed for a similar timeout bug
02
What the change forgot
Mid-edit, before any commit message, it names the file that usually changes with yours and is missing from the diff.
e.g. billing/ledger.py · changes with retry.py 9 of 11 times
03
What still holds the old name
Rename or remove an identifier or a literal, and it lists the files that still hold the old one.
e.g. config/settings.py · still references SESSION_TTL
04
Tests to update
For any change, the test files that will need edits too, ranked, so the pull request arrives complete.
e.g. tests/test_retry.py · moves with retry.py
05
Code health that learns
A health signal whose weights are learned from how real projects needed fixes, rather than set by hand.
e.g. payments/processor.py · first in line for review
06
Confidence you can act on
Every answer carries a calibrated band and the evidence behind it, so an agent or a reviewer knows when to trust it.
e.g. likely · 3 signals agree
History becomes prediction
Finds the file that keyword search misses.
Small models trained once on the history of other open-source projects, then used on any codebase. No LLM, milliseconds, on a CPU.
87%
Right file in the top five
2.6x
Better where search fails
3 in 4
Right file in the top five on real bugs
6 of 7
Works on code it never saw
Trained by us
Foresight is a set of machine-learning models we trained ourselves, on years of real commits, fixes and reviews across many open-source projects. They learned what usually happens next in a codebase, and apply it to yours with no training on your code.
Keeps learning
We retrain the models as new history comes in, so each release learns from more real changes than the last.
- Under half a second on a laptop
- Fully local, air-gap ready
- Explains every prediction
- Trained on open-source history, never on your code
Results on real bugs, each run on the code as it was before the fix, on projects the models never trained on.
How it runs
Built to run where your code lives.
Small models, trained once on the history of many open-source projects, then run on yours. Nothing trains on your code.
- No LLM
- Small gradient-boosted models. No prompt, no tokens, the same answer every time.
- CPU only
- Runs on the machine that already runs Repowise. No GPU.
- Milliseconds
- Fast enough to run on every commit and every pull request.
- Air-gap ready
- Inference is local. Self-hosted and air-gapped deployments keep code inside your network.
- Reads your history
- Each answer uses your repository's own co-change, fix and ownership record.
- Evidence behind every answer
- Each ranked file comes with the reason it ranked, in plain words.
Waitlist
Be first to run Foresight on your code.
Foresight is coming to Repowise hosted and enterprise. Leave a work email and we will bring teams in as it opens.
Rather see it on a call? Book a demoFREQUENTLY ASKED
Before you join
When is Foresight available?
Foresight is coming to Repowise hosted and enterprise. Join the waitlist and we will write to you when it opens for your team, or book a demo to talk it through with us.
Is Foresight open source?
No. The Repowise engine is open source under AGPL-3.0 and stays complete without Foresight. Foresight is the learned layer on top, and it is part of Repowise hosted and enterprise.
Does my code leave my network?
Foresight makes no LLM call, so nothing is sent to a model provider. The models are trained ahead of time on public open-source history and run on a CPU next to the index, which lets enterprise deployments run it self-hosted or fully air-gapped. It does not train on your code.
How is it different from the open-source index?
The open-source index is deterministic analysis: it tells you what is true about your code today, from the call graph, history, health and decisions. Foresight is learned. Its trained models read that same evidence to predict what a piece of work will need next: where a fix goes, what a change forgot, and which tests to update.
Does it replace my coding agent?
No. Foresight answers reach your agent and your reviewers through the same Repowise MCP tools, pull request checks and dashboard you already use, so the agent starts from the right files.