No — this is a build-to-order module, not an instant apps.odoo.com download. After a short discovery call ECOSIRE designs, builds, installs and tunes the assistant against your actual catalog, knowledge base and brand voice. Typical delivery is 2–4 weeks depending on catalog size, number of languages, and how deep you want the guardrails and analytics. We usually stand it up on a staging copy first, get your sign-off, then go live on production.

You receive the full module source in a Git repo you own, plus documentation and a training session, so you are never locked to us. Every engagement includes a post-go-live support and bug-fix window (scope agreed up front) covering defects, prompt/guardrail tuning and LLM-provider config help. Beyond that we offer optional ongoing support and version-upgrade plans — including migrating the addon cleanly when you move between Odoo 17, 18 and 19.

The module is built with a provider layer so you can use OpenAI GPT-4o, Google Gemini, Groq (Llama/Mixtral), or a self-hosted open-source model — chosen in Odoo Settings and swappable without a rebuild. It runs on Odoo 17, 18 and 19, Community or Enterprise, built as a portable addon so future upgrades are a controlled migration rather than a rewrite.

No — that's the core design. The assistant uses a retrieval-augmented pattern: it answers only from context retrieved live from your Odoo ORM (product.template, product.product, stock, pricelists, knowledge.article), and it's constrained by a system prompt and allow/deny guardrails. It won't recommend unpublished or archived SKUs, won't override pricing, and refuses off-catalog requests instead of hallucinating. Policy answers cite the source knowledge article.

Qualified shoppers are written into crm.lead as first-class Odoo records with the transcript, product interest and a computed lead-quality score, so they flow through your normal sales pipeline. Conversations are stored in dedicated models (ai.chat.session / ai.chat.message) governed by ir.model.access.csv and record rules — agents only see their team's chats. LLM API keys are stored server-side and never exposed to the browser; only the context needed to answer a question is sent to your chosen provider, and self-hosted models keep everything in your environment.

Yes. When the model hits a guardrail or the shopper asks for a person, the session escalates through an automated action (base.automation) — routing to your Odoo Live Chat, opening a helpdesk.ticket, or assigning the CRM lead to a salesperson, whichever fits your flow. The full transcript travels with the handoff so the human agent has instant context.

To ground the assistant we work against a copy of your catalog and knowledge base, typically on a staging instance you provision or grant us access to. We don't need your production credentials to design and test; final installation is done on production with your team present. All access and scope are agreed in writing before work starts.