Docs
/
Architecture
/
20260425 concept review
Introduction
Previous
20260503 repo split decision
Next
On this page
1. Executive Summary
2. Conceptual Model — Walkthrough
2.1 Entity Relationship Diagram
2.2 Model Providers
2.3 MCP Tools (External)
2.4 Internal Tools
2.5 Agents (Agent Profiles)
2.6 Job Definitions (ScheduledJob)
2.7 Job Runs
2.8 Channels
3. Validate Bruno's Mental Model
Statement 1: "We have a collection of available LLMs models in the Model Provider section"
Statement 2: "We have a collection of MCP tools"
Statement 3: "We have a collection of internal tools"
Statement 4: "Each agent will use a model provider, and can choose which tools may use"
Statement 5: "Each agent can also choose which MCP tools will be used"
Statement 6: "Note: explain how the tool approval system works here?"
Statement 7: "We have a collection of job definitions"
Statement 8: "Each job is a prompt action that will use an Agent, and will run in different modes: run-once, scheduled, webhook"
Statement 9: "Question: Is the job definition supposed to have a state active, disabled, cancelled, deprecated?"
Statement 10: "Should we also keep an history of the job definition state change logs?"
Statement 11: "We have a collection of job runs, that represents the total execution of a job run"
Statement 12: "Note: how the chat feature applies here? Can be considered a job run?"
Statement 13: "Each job definition is also considered a channel, so in the channel website we can browse and see the jobs history and more details"
4. Open Questions — Bruno's Direct Questions
4a. Tool Approval System
4b. Job Definition State Machine
4c. Chat as Job Run?
4.5 Recommendations Adopted (Implementation Plan)
5. Improvement Opportunities
🔒 Security
✨ UX
🎤 Demo Storytelling
🧪 Tests
📚 Docs
6. Recommended Next Sprint (5–7 Issues)
Appendix: Key File References
Scroll to top
Ask AI about this page
Copy as Markdown
Was this page helpful?
Yes
No