vrr/Pulse
2
0
Fork 0
mirror of https://github.com/rcourtman/Pulse.git synced 2026-04-28 03:20:11 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

docs: add comprehensive deep-dive documentation for AI subsystems

Browse source 
Adds detailed architecture documentation for Pulse Patrol and Pulse Assistant. Updates AI.md and PULSE_PRO.md. Also includes additional tests.
This commit is contained in:
rcourtman 2026-02-02 10:29:07 +00:00
parent df43f08cf2
commit fa1b74792e
22 changed files with 4322 additions and 203 deletions
Download patch file Download diff file Expand all files Collapse all files

45
docs/PULSE_PRO.md
View file

@ -40,22 +40,53 @@ Scheduled background analysis that correlates live state + metrics history to pr
### Pro-Only Automations
- **Alert-triggered analysis**: on-demand deep analysis when alerts fire.
- **Autonomous mode**: optional diagnostic/fix commands through connected agents.
- **Auto-fix**: guarded remediations when enabled.
- **Kubernetes AI analysis**: deep cluster analysis beyond basic monitoring (Pro-only).
- **Auto-fix mode**: automatic remediation with verification loops (see Autonomy Levels below).
- **Full autonomy unlock**: auto-fix for critical findings without requiring approval.
- **Kubernetes AI analysis**: deep cluster analysis beyond basic monitoring.
- **Audit-triggered webhooks**: real-time delivery of security events to external systems.
- **Advanced Reporting**: scheduled or on-demand PDF/CSV infrastructure health reports.
- **Agent Profiles**: centralized configuration profiles for fleets of agents.
### Autonomy Levels
Patrol and the Assistant support tiered autonomy:
| Mode | Behavior | License |
|------|----------|--------|
| **Monitor** | Detect issues only. No investigation or fixes. | Free (BYOK) |
| **Investigate** | Investigates findings and proposes fixes. All fixes require approval. | Free (BYOK) |
| **Auto-fix** | Automatically fixes issues and verifies. Critical findings require approval by default. | **Pro** |
| **Full autonomy** | Auto-fix for all findings including critical, without approval. | **Pro** (explicit toggle) |
### Investigation Orchestration
When Patrol creates a finding, the investigation orchestrator can:
1. **Create a chat session** dedicated to the finding.
2. **AI analyzes** the issue using available tools (metrics, logs, storage, etc.).
3. **Propose a fix** with risk assessment (low/medium/high/critical).
4. **Queue for approval** or **auto-execute** based on autonomy level.
5. **Verify the fix** with a follow-up read after execution.
Investigation outcomes include:
- `resolved` — Issue resolved during investigation
- `fix_queued` — Fix proposed, awaiting approval
- `fix_executed` — Fix auto-executed successfully
- `fix_verified` — Fix worked, issue confirmed resolved
- `needs_attention` — Requires human intervention
- `cannot_fix` — Issue cannot be automatically fixed
### What Free Users Still Get
- **Pulse Patrol (BYOK)**: background findings with your own provider.
- **Pulse Patrol (BYOK)**: background findings and investigation proposals with your own provider.
- **AI Chat (BYOK)**: interactive troubleshooting with your own API keys.
- **Update alerts**: container/package update signals remain available in the free tier.
### What You See In The UI
- **Patrol findings**: a prioritized list with severity, evidence, and recommended fixes.
- **Investigation status**: progress indicators showing investigation state and outcome.
- **Approval cards**: pending fixes await your review with one-click approve/deny.
- **Alert timelines**: AI analysis events attached to the alert history for auditability.
- **Remediation controls**: explicit toggles for autonomous mode and auto-fix workflows.
- **Remediation controls**: explicit toggles for autonomy mode in Patrol settings.
- **Agent profiles**: create, edit, and assign profiles in **Settings → Agents → Agent Profiles**.
## Pro Feature Gates (License-Enforced)
@ -112,6 +143,10 @@ This returns a feature map like `ai_alerts`, `ai_autofix`, and `kubernetes_ai` s
- **Alert-triggered analysis**: runs per alert event and writes analysis into the alert timeline for auditability.
- **Auto-fix safety**: requires explicit toggles and uses the same agent command scopes you configure for manual runs.
📖 **For complete technical details on the AI subsystems:**
- [Pulse Patrol Deep Dive](architecture/pulse-patrol-deep-dive.md) — Baseline learning, pattern detection, forecasting, correlation analysis, incident memory
- [Pulse Assistant Deep Dive](architecture/pulse-assistant-deep-dive.md) — Context prefetching, FSM enforcement, knowledge accumulation, safety gates
## Example Finding Payload (API)
`GET /api/ai/patrol/findings` returns structured findings you can integrate with external tooling: