SignalDesk
A local-first customer feedback intelligence tool that accepts CSV feedback data and outputs a risk-ranked customer list, customer segments, satisfaction signals, and a retention brief. It uses a layered architecture combining a transparent baseline with optional local models.
Timeline (last 90 days)
SignalDesk is an entry in the "side hustle" theme of week two of the Hacktoberfest open source AI challenge.
SignalDesk is positioned as a local tool that turns a CSV into a risk-ranked customer list, segments, satisfaction and wait-time signals, and a retention brief.
The current version of SignalDesk is started via `python app.py` and used by visiting `127.0.0.1:8765`.
The first version of SignalDesk is a small web app using only the Python standard library.
SignalDesk computes a transparent local risk baseline, aggregates results by segment and generates retention briefs, with every decision traceable.
SignalDesk calls the Gemma model via Ollama to generate briefs; when Ollama or the model is unavailable, a deterministic fallback takes over and the core workflow remains usable.
SignalDesk reserves an adapter seam for TabPFN, but full integration is not yet complete.
SignalDesk clearly labels which model produced a brief, making it easy for enterprises to trace the source of conclusions in compliance or audit scenarios.
The author explicitly states that no paying customers have been acquired yet.
The project applied for the Best Use of Gemma and Best Use of TabPFN categories, but the author states that partner categories will only be formally claimed once the corresponding technologies have been verified in the submitted build.
1 more timeline events
All Facts (11)
The project applied for the Best Use of Gemma and Best Use of TabPFN categories, but the author states that partner categories will only be formally claimed once the corresponding technologies have been verified in the submitted build.
50%UnverifiedSignalDesk is positioned as a local tool that turns a CSV into a risk-ranked customer list, segments, satisfaction and wait-time signals, and a retention brief.
50%UnverifiedThe current version of SignalDesk is started via `python app.py` and used by visiting `127.0.0.1:8765`.
50%UnverifiedThe first version of SignalDesk is a small web app using only the Python standard library.
50%UnverifiedSignalDesk computes a transparent local risk baseline, aggregates results by segment and generates retention briefs, with every decision traceable.
50%UnverifiedSignalDesk calls the Gemma model via Ollama to generate briefs; when Ollama or the model is unavailable, a deterministic fallback takes over and the core workflow remains usable.
50%UnverifiedSignalDesk reserves an adapter seam for TabPFN, but full integration is not yet complete.
50%UnverifiedSignalDesk clearly labels which model produced a brief, making it easy for enterprises to trace the source of conclusions in compliance or audit scenarios.
50%UnverifiedThe author explicitly states that no paying customers have been acquired yet.
50%UnverifiedThe first validation target is a small service business that still relies on manual review of customer feedback; the proposed pilot plan is to first run weekly reports on synthetic data, then do a privacy-preserving local run using the business owner's own export data.
50%UnverifiedSignalDesk is an entry in the "side hustle" theme of week two of the Hacktoberfest open source AI challenge.
50%