AI / Search
Convolations
AvailableA self-hostable conversation-log search stack: ingest JSON/CSV/TXT (and optional email), index with Meilisearch and Qdrant, then expand context around a hit. Built for archives that should stay on your machines.
A team with large chat or ops logs gets a retrieval product they can run themselves — keyword plus vectors — without sending the corpus to a third-party workspace.
Build vs buy
Docker Compose, dual indexes, ingestion parsers, and context expansion via sequential IDs are already designed. The buyer is not starting at “which vector DB.”
Differentiation
Dual keyword + semantic search with fast context windows, packaged as a compose stack rather than a notebook demo.
Best fit
Teams that accumulate large chat/ops logs and need searchable retrieval without sending everything to a hosted SaaS.
Evidence
Screenshots are the working interface — not a pitch deck. Click to enlarge.
Click any image to open the gallery · 3 images. Use arrow keys in the lightbox.
Why it matters
Conversation archives are usually dead weight. Convolations treats them as a corpus you can actually query — privately.
What it does
Operators ingest logs, index them for keyword and vector search, then expand context around matches for investigation or reuse.
Operational advantage
Avoids both a custom RAG project and a hosted “AI notes” subscription for material that is already sitting in exports.
Build vs buy
Docker Compose, dual indexes, ingestion parsers, and context expansion via sequential IDs are already designed. The buyer is not starting at “which vector DB.”
Interface advantage
Click a hit and the surrounding messages open immediately. The product is built around investigation, not a chat toy.
What makes this different
Dual keyword + semantic search with fast context windows, packaged as a compose stack rather than a notebook demo.
Where this gets used
- Search a year of Cursor or ChatGPT exports
- Investigate an ops incident across chat logs
- Keep retrieval in-house for confidentiality
Commercial paths
- Self-hosted license for agencies and internal ops teams
- Managed hosting upsell around the same stack
- Vertical pack for support logs or engineering transcripts
Best buyer fit
Teams that accumulate large chat/ops logs and need searchable retrieval without sending everything to a hosted SaaS.
Current status
Available for licensing or exclusive acquisition. UI is a working Streamlit surface, not a multi-tenant SaaS console.
Key capabilities
- Meilisearch keyword search and Qdrant semantic search
- Embeddings with local/open-source model support
- Context expansion around results
- JSON/CSV/TXT ingestion
- Docker Compose deployment
Related inventory
Business Software
The Practical Effect
Sell structured business reviews with intake, Stripe checkout, and a 30/60/90 measurement loop already wired.
Business Software
ForgeSales
Run human and agent outreach on one pipeline instead of letting drafts and CRM notes live in separate tools.
AI / Search
NicheScout AI
Get a first pass of niche businesses and contact surfaces from a research UI instead of a morning of manual search.