This release adds 3 notable features for engineering teams evaluating rollout.
✓ No known CVEs patched in this version
Topics
+4 more
Summary
AI summaryUniversal agent adapter conformance and native persistent session ledger introduced.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Feature | Medium |
Adds universal agent adapter conformance for multiple LLM profiles. Adds universal agent adapter conformance for multiple LLM profiles. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Introduces `contextlattice-agent-adapter` lifecycle helper for bootstrap, context pack, checkpoint, handoff, event, and completion flows. Introduces `contextlattice-agent-adapter` lifecycle helper for bootstrap, context pack, checkpoint, handoff, event, and completion flows. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Adds native agent runtime sessions with persistent ledger, events, telemetry, and Agent Radar dashboard data. Adds native agent runtime sessions with persistent ledger, events, telemetry, and Agent Radar dashboard data. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Implements objective runtime contract output with bounded evidence, delta, risk/blocker, and next-action state. Implements objective runtime contract output with bounded evidence, delta, risk/blocker, and next-action state. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Adds auto-session ensure/recover across agent start, policy pack, context pack, Dream Mode, writeback, and compaction hooks. Adds auto-session ensure/recover across agent start, policy pack, context pack, Dream Mode, writeback, and compaction hooks. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Medium |
Hardens hosted Caddy ingress with h1 default, audited h2 override, and memory/PID caps. Hardens hosted Caddy ingress with h1 default, audited h2 override, and memory/PID caps. Source: llm_adapter@2026-06-05 Confidence: low |
— |
| Feature | Low |
Adds public memory-spine story, local AI workspace comparison, scaling‑memory page, and minimal click/tap suite cards. Adds public memory-spine story, local AI workspace comparison, scaling‑memory page, and minimal click/tap suite cards. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Feature | Low |
Adds `scripts/agent/audit-universal-agent-adapter` to validate offline and live lifecycle conformance. Adds `scripts/agent/audit-universal-agent-adapter` to validate offline and live lifecycle conformance. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Refactor | Low |
Updates `config/agents/agent_profiles.json` to define supported agent profiles. Updates `config/agents/agent_profiles.json` to define supported agent profiles. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Refactor | Low |
Updates `config/agent_contracts/agent_output_contracts.json` to include universal adapter and objective‑runtime contracts. Updates `config/agent_contracts/agent_output_contracts.json` to include universal adapter and objective‑runtime contracts. Source: llm_adapter@2026-06-05 Confidence: high |
— |
| Refactor | Low |
Adds `scripts/agent/audit-agent-runtime-contract` for validating objective runtime and auto‑session behavior. Adds `scripts/agent/audit-agent-runtime-contract` for validating objective runtime and auto‑session behavior. Source: granite4.1:30b@2026-06-05-audit Confidence: low |
— |
Full changelog
ContextLattice v3.3.45 turns the recent agent-memory work into a real runtime contract release.
Highlights:
- Universal agent adapter conformance for Codex, Claude Code, OpenCode, Hermes, Pi, ChatGPT web/desktop, and Claude web/desktop profiles.
- New
contextlattice-agent-adapterlifecycle helper for bootstrap, context pack, checkpoint, handoff, event, and completion flows. - Native agent runtime sessions with persistent session ledger, session events, runtime telemetry, and Agent Radar dashboard data.
- Objective runtime contract output with bounded evidence, objective delta, risk/blocker, and next-action state.
- Auto-session ensure/recover across agent start, policy pack, context pack, Dream Mode, writeback, and compaction hooks.
- Hosted Caddy ingress hardening: h1 default, audited h2 override, and Caddy memory/PID caps.
- Public memory-spine story, local AI workspace comparison, scaling-memory page, and minimal click/tap suite cards.
- Public boundary cleanup: internal planning/doc references remain out of public-facing docs.
For agent developers:
config/agents/agent_profiles.jsondefines supported profiles.config/agent_contracts/agent_output_contracts.jsonnow includes universal adapter and objective-runtime contracts.scripts/agent/audit-universal-agent-adaptervalidates offline and live lifecycle conformance.scripts/agent/audit-agent-runtime-contractvalidates objective runtime and auto-session behavior.
Verification summary:
- Universal adapter audit passed for 10 configured profiles with no findings.
- Runtime contract audit passed with objective_runtime_status=passed for preflight and CLI paths.
- Git whitespace check passed before tagging.
- Live local gateway health was OK before release.
- Agent runtime telemetry after stale audit cleanup: active=3, completed=55, failed=0, blocked=0, total=58.
Lane companion tags:
v3.3.45-public-> public/main858051fv3.3.45-origin-> private origin/main1ddfb5fv3.3.45-public-paid-> public-paid/main5793a15
Notes:
- The remaining active runtime sessions are not UI elements; they are real persisted agent sessions from current/adjacent agent workflows.
- Fourteen stale objective-runtime audit/preflight sessions from before the audit cleanup were completed before this release.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About sheawinkler/ContextLattice
Private-by-default memory and context layer for agents with Go/Rust runtime, staged retrieval across fused data backends, and long-horizon context continuity.
Related context
Related tools
Beta — feedback welcome: [email protected]