AI Natural-Language Data Assistant (Ask Your Odoo)

This is a build-to-order module with a typical lead time of 2–4 weeks depending on how many models and question types you need. The process is: a short discovery workshop to map your models, vocabulary, and priority questions; ECOSIRE builds the query catalog, OWL/QWeb interface, security rules, and LLM adapter; we install on your staging instance for review; then we deploy to production and hand over docs and training. Larger catalogs or heavy customization may extend the timeline, and we confirm the exact scope and date before we start.

Every engagement includes a post-go-live support window for bug fixes and query-catalog tuning as your team discovers new questions they want to ask. We keep the module compatible with your Odoo edition and version during that window, and we provide a clear runbook and (optionally) an ongoing support/maintenance retainer for adding new question types, upgrading across Odoo major versions, or extending the catalog to new modules over time.

No. By design the assistant never generates or runs SQL. ECOSIRE builds a catalog of pre-written, parameterized queries using the Odoo ORM (search/read_group), and the LLM's only role is to pick the right catalogued query and extract safe parameters like date ranges and team names. This eliminates prompt-injection-to-SQL-injection risk entirely and follows Odoo's rule of never using raw SQL on untrusted input. The assistant is also read-only — it cannot create, update, or delete records.

Yes. Because every catalogued query runs through the Odoo ORM as the asking user, your ir.model.access.csv access rights and record rules apply automatically — multi-company boundaries, sales-team ownership, and any per-record restrictions. A salesperson only sees their own pipeline; finance data stays behind finance access. There is no bypass path, because the assistant uses the exact same security layer as the rest of Odoo.

We build and deliver on Odoo 17, 18, and 19, and it works on both Community and Enterprise. The assistant relies on standard framework features — the ORM, security model, OWL/QWeb UI, and optionally automated actions — so it does not require Enterprise. If you're on Community we tailor the charting and UI to what's available; if you're on Enterprise we can lean on additional built-in components where useful.

Only the natural-language question text and the catalog's query descriptions (schema/labels) are sent to the LLM so it can classify intent — your actual record data is not sent to the model. The query itself then runs locally in Odoo through the ORM. If your policy requires it, ECOSIRE can wire the adapter to a self-hosted or private model so nothing leaves your infrastructure. API keys are stored in Odoo system parameters, and every question is captured in the audit log.

Yes. The query catalog is designed to grow. ECOSIRE hands over documentation on how catalogued queries are structured, and each new question type is a reviewed, parameterized query we (or a trained admin) add and map to plain-English phrasings. This keeps every addition safe and auditable — new capability never means new free-form SQL. Catalog expansion is a common part of the post-launch support window or an ongoing retainer.