buildPatrolAssessmentAssistantHandoff lost its last caller when the Patrol
summary card's Discuss handler was removed in eac7dfe9ef (2026-05-15), and
buildPatrolConfigurationFailureHandoff lost its last caller in 27ba85ddc6
(2026-06-23). Since then only their own tests called them. A TypeScript
compiler import graph rooted at src/index.tsx confirms no reachable file
imports either one. Removing them also drops PatrolInvestigationContextSummary,
their input and output types, and the private assessment, recent-change,
correlation and save-failure helpers that only they used (found with
noUnusedLocals until it reported nothing).
Rollup had already tree-shaken all of it: production builds of the parent
tree and of this tree are byte-identical across all 187 dist files, and the
bundle-size check passes. The presentation helpers the dead code imported
keep their live callers.
Their test cases go with them. Two cases covered branches the live finding
handoff still uses, so they now run through
buildPatrolAssistantFindingHandoffActions and buildPatrolAssistantFindingHandoff:
a typed pending action without a legacy approval id stays approval-bound,
and the finding context qualifies its resource with the record's node and
truncates long lines. v8 coverage of the surviving model code matches the
pre-removal baseline except one defensive guard the live caller cannot
reach. The raw-source guards pinned none of the removed code.
The Assistant drawer still restores saved patrol_assessment and
patrol_configuration_failure sessions, so that compatibility path stays.
The contracts described both handoffs as live producers, and described a
Patrol summary card, assessment strip, assessment readout, score chip and
header trust line that the page no longer renders. patrol-intelligence,
ai-runtime, api-contracts, frontend-primitives, security-privacy,
storage-recovery, agent-lifecycle and unified-resources now say what the
code does: those handoffs survive only as saved sessions,
the drawer, session list and API keep restoring them, rules for a returning
summary card, strip or readout are conditional, and handoff_actions and
adjacent-boundary clauses name the finding and attention handoffs that send
them today. The run-history handoff clauses also claimed the browser sends a
[Patrol Run Context] block and scoped resources; it sends the briefing and
run identity, and internal/ai/patrol_assistant_handoff.go rebuilds the rest
from the run ID.
docs/AI.md (and its public mirror) told users the legacy operational score
sits in a collapsed supporting-context section. That section and the score
are gone; the section now points at Activity > Finding options and history,
where check history lives. ai_runtime_docs_policy_test.py, its registered
proof, now pins that location and rejects the retired claim. The same file
had failed on main since 2789a15fe5 reworded the Patrol mode API sentence to
"The API retains `autonomy_level` for compatibility" without updating its
pin; the pin now matches that wording.
Explain the missing v6.4.5 undo control and the existing scoped API. Verify exact selection, confirmation, browser session and CSRF, failure handling and readback without copying credentials or deleting unrelated rules.
Contract-Neutral: Documentation and regression-proof coverage of existing Patrol suppression APIs only; no runtime, route, authority or subsystem contract changes.
Change-source: pulse-maintainer
AI.md, AI_AUTONOMY.md and PULSE_PRO.md linked nine times into
docs/architecture/, which .gitignore marks as not for public release. The
targets were never missing, they were deliberately unpublished, so every one
of those links was dead for anybody but the maintainer.
Write the three promised documents against the code and publish them under
docs/ where the rest of the shipped set lives.
PATROL_ARCHITECTURE.md covers a run end to end. The interesting part is that
deterministic signal detection runs after the model, not before, so unmatched
signals catch what the model failed to file. Signal types, the thresholds
those signals derive from your own alert settings, and every condition in
Finding.ShouldInvestigate are documented from internal/ai/patrol_signals.go
and internal/ai/findings.go.
ASSISTANT_SAFETY.md documents the session state machine in
internal/ai/chat/fsm.go, its four states, the tool classification it runs on,
and its invariants. No write without a validated target, no second write
before the first is verified, no final answer about an unverified change, and
no attempt count that wears the gate down.
ASSISTANT_ARCHITECTURE.md covers the agentic loop around that machine, the
three-phase pipeline, why only execution parallelises and at what cap, the
read-before-write batch that must stay ordered, the look-before-asking gate
and its bound, and the stable error codes.
Also removed two older pointers into the same private directory, in API.md
and UPGRADE_v6.md, and the two references to ENTITLEMENT_MATRIX.md in
PULSE_PRO.md. That file exists locally and is a monetization document, so
publishing it is not a documentation decision.
Unresolvable intra-doc links are now 2 of 218, both internal release-control
documents deliberately withheld from the shipped set.
Contract-Neutral: documentation only
The readiness advisor failed capable local models for adapter and probe
defects rather than model incapability (#1624 Ollama, #1614 llama.cpp):
- Send num_ctx sized to the haystack fixtures (clamped to the model's
trained window) so Ollama no longer truncates ~25KB prompts at its
4096-token server default; the trained-window guard alone passed while
the runtime request was being truncated.
- Forward an explicitly pinned temperature 0 instead of dropping it to
Ollama's 0.8 default against a nonce-exact validator (ChatRequest gains
TemperatureSet; Ollama options temperature is now a pointer).
- Raise the probe generation cap from 256 to 2048 tokens so qwen3-style
<think> reasoning cannot exhaust the budget before the tool call, and
surface the provider done_reason when validation fails.
- Synthesise tool-call IDs in the OpenAI-compatible adapter (streaming
finalizer and buffered path) when the server omits them, as llama.cpp
commonly does, mirroring the Ollama adapter instead of failing tool
protocol 0/3 on transport shape.
- Probe with the Patrol loop's 60s stream stall allowance instead of the
12s chat default (chat.PatrolProviderStreamIdleTimeout is now exported).
- Stop discarding probe and validator errors: log them, carry them in a
new PatrolModelReadinessResult.Details field surfaced through the API
snapshot and Settings UI, and keep a transport-level probe failure's
specific diagnosis instead of overwriting it with the generic
capability wording.
Builds on 4a2335ce7, which already reclassifies protocol failure as
"provider connected; Patrol capability not verified".
Fixes#1624Fixes#1614
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Manifest-backed MCP tools, prompts, and resources with surface affordance contracts; agent capability manifest and governance projection; API contract tests and capability route projection; operations-loop and intelligence-funnel telemetry; release-control subsystem documentation, registry, and tooling; licensing and configuration.
- Major updates to README.md and docs/README.md for Pulse v5
- Added technical deep-dives for Pulse Pro (docs/PULSE_PRO.md) and AI Patrol (docs/AI.md)
- Updated Prometheus metrics documentation and Helm schema for metrics separation
- Refreshed security, installation, and deployment documentation for unified agent models
- Cleaned up legacy summary files
The storage deduplication logic only checked cluster config's Shared
flag, but this required the cluster config API call to succeed. When
the per-node storage API already returns shared=1 (as the user
verified), we should use that directly.
Now we check three sources for shared storage detection:
1. Per-node API shared flag (storage.Shared)
2. Cluster config shared flag (if available)
3. Storage type heuristics (NFS, RBD, PBS, etc.)
Related to #1049
Add contextual help icons throughout the UI to improve feature
discoverability. Users can click (?) icons to see explanations
with examples for settings they might not understand.
- HelpIcon component with click-to-open popover
- Centralized help content registry in /content/help/
- FeatureTip component for dismissible contextual tips
- Help added to: alert delay, AI endpoints, update channel
- Add comprehensive explanation of what data Patrol receives
- Document the enriched context (trends, baselines, predictions)
- Explain operational memory (notes, dismissed alerts, incidents)
- Clarify why Patrol catches issues that static alerts miss
- Mark Patrol as Pro feature with link to pulserelay.pro
- Add validateAIFindings() that cross-checks AI findings against actual metrics
- Filter out low-confidence findings (CPU <50%, memory <60%, disk <70%)
- Always allow critical findings, backup issues, and reliability findings through
- Update AI system prompt with stricter thresholds and explicit noise examples
- Add 'before creating a finding' checklist for AI (the 3am test)
- Update AI.md docs with clear value proposition and expectations
- Add comprehensive tests for the validation layer
This ensures paying users get immediate value without noise.
Adds IncludeAllDeployments option to show all deployments, not just
problem ones (where replicas don't match desired). This provides parity
with the existing --kube-include-all-pods flag.
- Add IncludeAllDeployments to kubernetesagent.Config
- Add --kube-include-all-deployments flag and PULSE_KUBE_INCLUDE_ALL_DEPLOYMENTS env var
- Update collectDeployments to respect the new flag
- Add test for IncludeAllDeployments functionality
- Update UNIFIED_AGENT.md documentation
Addresses feedback from PR #855
rcourtman