AI Discuss (BytesRaw)
Chat with AI agents natively in Discuss, in separate conversations
Your AI agents,
in the messaging menu.
AI Discuss puts your AI agents where your team already talks. A new AI tab shows the agents you are allowed to use as a grid of avatars — click one to carry on where you left off, or click its (+) badge to start a fresh conversation. Below a divider, the same tab lists your existing AI chats.
Conversations are real Discuss channels, so unread counters, previews, the typing indicator and full history all behave the way messaging already does. But they stay in their own lane — AI chats appear in the AI tab only, never mixed into your notifications — and each agent replies through a hidden partner, so your Contacts stay clean.
Chat Where You Already Are
No separate app, no new window to learn. The AI tab sits next to Notifications, Chats and Channels in the messaging menu, and each conversation is an ordinary Discuss thread you can scroll, search and come back to.
The Agent Sees What You See
Turn on Use Active Record and starting a chat from a record hands the agent that record — or your whole multi-record selection — as context, and titles the conversation after it. “Who is this customer?” just works, with no copying and pasting.
It Answers As You, Not As Admin
Every tool the agent runs during a conversation executes with your access rights. If you cannot see a record in the interface, the agent cannot read it for you either — and each call is written to the tool audit log under your name.
A chat surface,
built for agents.
Thirteen behaviours that make an AI conversation feel like a native part of Discuss rather than a bolted-on chatbot.
A Dedicated AI Tab
The messaging menu gains an AI filter alongside Notifications, Chats and Channels. It shows every agent available to you as an avatar launcher, then your AI conversations, with a Show all conversations link once the list grows past what fits.
Separate Conversations, on Purpose
Clicking an agent resumes your most recent usable chat with it; the (+) badge always starts a new one. A chat is only resumed when it would open with the same context and has not gone stale, so an agent anchored to a record never silently answers about the previous one.
Conversations That Name Themselves
A brand-new chat is retitled from your opening question — or from the file you dropped in — so the conversation list reads as a list of topics rather than a stack of identical agent names. A title you set yourself is never overwritten.
Hidden Bot Partners
Each agent authors its replies through an archived partner created on first use, carrying the agent’s own avatar. Replies look like messages from a real participant without a single extra row appearing in Contacts.
Send Files to the Agent
Attach an image, a PDF or a voice note and the agent handles it: files go to the model natively where the provider supports it, text files are inlined, and audio is transcribed first. A message with nothing but a file attached is still answered.
Long Conversations Without Losing the Thread
The most recent turns are sent to the model verbatim; everything that scrolls out of that window is folded into a rolling summary refreshed in batches rather than on every turn. Long chats keep their context without resending the entire transcript each time.
Expiry Instead of Stale Context
When a conversation has been idle past the agent’s expiry window, the next message gets a polite nudge to start a new chat rather than an answer built on week-old context. The chat stays fully readable, and abandoned ones are cleaned up automatically.
Confirmations in the Flow of Chat
When an agent wants to run a tool that changes something, it describes what it intends to do and waits. Your next message approving it is what lets the tool run — the confirmation state is remembered on the channel between turns.
Markdown Answers, Rendered
The agent’s Markdown replies — headings, lists, tables, code, links — are converted to Discuss-friendly HTML, so answers are readable instead of arriving as raw asterisks.
Chatter Tools Included
Three ready-to-use tools ship with the module: Post Chatter Note logs an internal note on a record, Schedule Activity creates a to-do, call or meeting reminder, and Create Email Draft prepares an email a human must open and send. All three are confirmation-gated.
The “Odoo AI” Assistant, Ready to Go
A general-purpose agent ships with the module, with an Odoo Navigation skill that teaches it which view to open for which kind of question. It is read-only by construction: it can search, aggregate, read attachments, calculate and open the right screen — but holds no tools that create, modify or archive anything.
Per-Agent Visibility
An agent restricted to specific user groups only appears in the AI tab for members of those groups. Hide from Messaging Menu keeps a purpose-built agent out of the tab entirely while leaving it usable from the Agents view or a record’s chatter.
Chat From the Agent Record
The Agents kanban and form gain a Chat in Discuss action, so a manager can test an agent the moment it is configured, straight from the record they were editing.
Add a connection.
Start chatting.
The module works as soon as an agent has a provider connection; the rest is optional tuning.
Wire up the shipped agent
At AI → Agents → Odoo AI, add at least one line on the Connections tab.
Let it see the record
On the agent form, under Discuss, tick Use Active Record.
Control who sees what
Use Restricted To to limit an agent to user groups, and Hide from Messaging Menu to keep it out of the AI tab.
Set the conversation lifetime
Chat Expiry (days) decides when an idle conversation stops accepting new messages. Set 0 to never expire.
Add the chatter tools
On the agent’s Tools tab, add Post Chatter Note, Schedule Activity or Create Email Draft.
Tune history handling
Optional system parameters control rolling summarization and how long idle chats are kept before cleanup.
See it in action
The AI tab, a conversation in progress, and acting on a record from chat.
From first message to first action
Open a conversation, anchor it to a record, then let the agent do something about it.
Start Talking to an Agent
Open your first AI conversation and find your way back to it later.
Ask About the Record You Are On
Get answers about the specific quotation, contact or invoice in front of you, without describing it.
Let the Agent Act on a Record
Move from asking to doing, with a human approval in the middle.
Full transparency
The current limitations, stated plainly.
Very long chats are compressed, not remembered whole
History is fed to the model as a recent verbatim window plus a rolling summary of what came before. Summarization costs an extra provider call, batched so it does not run every turn — but a very long conversation is a compressed one, and fine detail from far back may be lost. Starting a fresh chat for a new topic gives better answers. Set ai_discuss_bytesraw.summarize_history to 0 to turn summarization off; older turns are then simply dropped.
Expired chats stop answering
Expiry is measured from the last message, not from creation, so an active chat never expires mid-use. Once expired, posting into it returns a nudge rather than an answer; the transcript stays readable.
Idle chats are cleaned up
AI chats that have gone idle are removed by a scheduled cleanup after the retention period (7 days by default). Rename or keep anything you need beyond that.
Questions, answered
Quick answers to the questions buyers ask most.
Do AI conversations clutter my Notifications tab?
No. AI chat channels are shown in the AI tab only.
Can the agent read data I am not allowed to see?
No. The agent loop runs with elevated rights only to read its own configuration and credentials; every tool it calls runs as you, so record rules and access rights apply exactly as they do in the interface.
Why does clicking an agent sometimes open an old conversation?
Clicking the avatar resumes your most recent usable chat with that agent, because reaching for an agent usually means carrying on. Use the (+) badge to force a new one.
Do the agent bots show up in my Contacts?
No. Each agent’s bot partner is created archived and exists only to author messages inside AI chats.
Can I stop an agent from appearing in the messaging menu?
Yes — tick Hide from Messaging Menu on the agent. It stays usable from the Agents kanban, a record’s chatter, or any other surface that owns it.
Does the shipped Odoo AI agent modify my data?
No. It is deliberately configured with read, introspection and navigation tools only. If you want it to change data, add write tools to it yourself.
Release history
A clean record of every release.
Bytesraw
🛠️ Need help?
Got a question, found a bug, or need a custom enhancement? Our support team responds within 24 hours on business days.
help@bytesraw.com →💼 Custom development
Need this module tailored to your workflow, or a brand-new Odoo solution? Let’s talk about your project.
info@bytesraw.com →| Availability |
Odoo Online
Odoo.sh
On Premise
|
| Odoo Apps Dependencies |
Discuss (mail)
|
| Community Apps Dependencies | Show |
| Lines of code | 6784 |
| Technical Name |
ai_discuss_bytesraw |
| License | OPL-1 |
| Website | https://www.bytesraw.com |
Odoo Proprietary License v1.0 This software and associated files (the "Software") may only be used (executed, modified, executed after modifications) if you have purchased a valid license from the authors, typically via Odoo Apps, or if you have received a written agreement from the authors of the Software (see the COPYRIGHT file). You may develop Odoo modules that use the Software as a library (typically by depending on it, importing it and using its resources), but without copying any source code or material from the Software. You may distribute those modules under the license of your choice, provided that this license is compatible with the terms of the Odoo Proprietary License (For example: LGPL, MIT, or proprietary licenses similar to this one). It is forbidden to publish, distribute, sublicense, or sell copies of the Software or modified copies of the Software. The above copyright notice and this permission notice must be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Please log in to comment on this module