Bring your data, lists, models and documents into decisions
You already pay for data providers and keep your own lists and database. Connect them, language models and documents to decisions using your own keys.
Connect data and services to decisions
Your keys stay in one vault.
Data APIs
Call any REST API or SOAP web service as part of a decision flow.
Lists
Rules check list values, and people can change entries without a new release.
Database
Read and write decisions, cases, outcomes, cached responses and list entries in your PostgreSQL database.
LLMs
Use a provider you choose or your OpenAI-compatible endpoint, with your key.
OCR
Send a document on a case to your configured OCR service and write its text to the case file.
Your model
Score in a decision flow with a model trained in Python, R or Julia.
Connect any HTTP API to a decision flow
- Connection optionsGET, POST, PUT or SOAP, with JSON or XML and Basic or OAuth2
- AI proposalAI proposes API settings for a person to review, apply and save
- Response cacheKeep provider responses in your database
Lists stay as data beside the versioned decision flow
- Live entriesPeople's changes appear on the next request, with no new release
- Add entriesAnalysts and systems can add entries to lists
- BacktestsCheck each past request against the list as it stood at the time
Your PostgreSQL or Supabase database holds decision data
- Decision archiveEach executed decision is written to a table in your database
- After each decisionLog its outcome. See rule performance. Only totals leave your database
- One connectionOne connection serves several destinations. Prepare creates their tables.
Call a language model from a decision flow or case
- Provider APIsOpenAI, Gemini, Anthropic Claude, Azure OpenAI, DeepSeek, Mistral
- Team promptThe team writes a prompt template filled from decision data
- Rules decideRules read the JSON reply under policy you set
OCR text joins the case file
- Document OCRA task sends an uploaded case document to your configured OCR service
- Case fileRecognised content is written there
- Decision logicAI tasks read the case file text, and decision logic decides
A model you trained scores in the decision flow
- Model importImport Python, R or Julia models. Declare Python version and libraries
- Input mappingThe team chooses which data value feeds each input
- Decision logicIt reads the model's score as a value in the flow
Keys live in one vault
Each data source, LLM model or OCR service holds only a reference, so its key stays out of configurations, exports and releases. This is part of Decisimo's security controls.
Failed calls return an error value
A failed or refused call to the public internet returns an error value in the data. The decision flow can route it to a fallback branch.
Requests reach the released decision
A released flow handles requests through an API, in the background or from a whole file.
API endpoint
JSON in and out, with its own API key
Webhook
The request is accepted at once and processed in the background
Batch
Run a whole file through a release to pre-score a portfolio
Questions we get asked
Can I connect a provider with an HTTP API?
Configure any REST API or SOAP web service as a data source.
Which language models can I connect, including my own?
Use OpenAI, Azure OpenAI, Google Gemini, Anthropic Claude, DeepSeek, Mistral or an OpenAI-compatible endpoint. You write the prompt, and rules read the JSON reply and decide.
Where are decisions and list entries stored?
Decision records and list entries are stored in the PostgreSQL database your organisation owns and controls. This can be a Supabase project.
What happens when a provider call fails?
A failed or refused call returns an error value in the data. The decision flow can route it to a fallback branch.
Who provides the data, LLM and OCR accounts?
You bring the account and key for each data provider, language model and OCR service. Credentials are held in the vault.