ur-agent 1.14.0 → 1.15.0

package/docs/AGENT_FEATURES.md

@@ -21,6 +21,11 @@ ur automation run nightly --dry-run
 ur automation run-due
 ur model-doctor
 ur a2a serve --dry-run
+ur bg run "fix the flaky parser test" --worktree --dry-run
+ur memory retention show
+ur code-index watch --dry-run
+ur ide diff capture --title "Working tree review"
+ur eval bench list
 ur semantic-memory build
 ur semantic-memory search "release checks"
 ur claim-ledger add --claim "..." --source web:https://example.com
@@ -29,6 +34,21 @@ ur browser-qa validate
 ur browser-qa run home-page-smoke --dry-run
 ```
 
+## v1.15.0 Additions
+
+Seven additions from the 2026 agent-platform gap list. They keep UR local-first
+and route model work through the local Ollama-backed UR runtime.
+
+| Addition | Surface | What it adds |
+| --- | --- | --- |
+| Managed background agents | `ur bg run|fanout|list|status|logs|attach|kill` | Detached, durable local agent runs with optional worktrees and opt-in PR creation through `gh` |
+| Auto compaction and memory retention | `compaction.autoThreshold`, `ur memory retention` | Configurable context compaction threshold plus TTL/max/decay pruning for `.ur/memory/*.jsonl` |
+| Code-index auto-reindex | `codeIndex.autoReindex`, `ur code-index watch` | File watcher that rebuilds the local semantic code index after source changes |
+| Live artifact steering | `ur artifacts comment <id> --feedback ... --task <bg_id>` | Artifact feedback is queued into the linked background task inbox and injected into active stream-json background agents as `priority: "now"` turns |
+| Full opt-in A2A task server | `ur a2a serve`, `/a2a/tasks` lifecycle routes | Token/delegation-protected task submission, status, output, and cancel routes backed by local background tasks |
+| IDE inline diff bundles | `ur ide diff capture|list|show|comment|schema`, `extensions/vscode-ur-inline-diffs/` | Editor-readable `.ur/ide/diffs/` manifest, metadata, patch files, comments, plus a native VS Code tree/webview review extension |
+| Benchmark adapters | `ur eval bench swe-bench|terminal-bench|aider-polyglot` | Imports local benchmark JSON/JSONL exports into UR eval suites without external downloads |
+
 ## Nine Points
 
 | Point | UR surface | What it adds |

package/docs/AGENT_TRENDS.md

@@ -16,11 +16,17 @@ ur agent-templates install
 ur model-doctor
 ur automation create nightly --schedule "0 9 * * 1-5" --prompt "Review open tasks"
 ur automation run-due
+ur bg run "fix the flaky parser test" --worktree --dry-run
+ur bg fanout "try two parser fixes" --agents 2 --dry-run
 ur agent-task pr --create --dry-run
 ur a2a serve --dry-run
 ur semantic-memory build
+ur memory retention show
 ur code-index build
+ur code-index watch --dry-run
 ur code-index search "where is the rate limiter configured"
+ur ide diff capture --title "Working tree review"
+ur eval bench list
 ur role-mode install all
 ur agent-task pr --create --dry-run   # runs the self-review gate first
 ur spec init checkout --goal "1. add cart 2. add payment 3. add receipt"
@@ -48,23 +54,24 @@ Inside an interactive session:
 | --- | --- | --- | --- |
 | Local-first model runtime | Covered | local Ollama app endpoint, local and Ollama Cloud-backed models exposed through that app, model auto-routing | Add model capability reporting for tools, vision, context length, and multimodal readiness |
 | MCP tool ecosystem | Covered | `ur mcp`, MCP OAuth/XAA helpers, elicitation, permission checks, shared tool registry | Keep MCP registry/security guidance current as the spec evolves |
-| A2A / Agent Card interoperability | Adapter-ready | `ur a2a card` and `/a2a-card` export Agent Card metadata | Add an opt-in A2A task server adapter for remote agent-to-agent task execution |
-| Durable workflows and checkpoints | Partial | resume, rewind, background tasks, cron/workflow internals, file restore | Publish a checkpointed workflow format for repeated automations |
+| A2A / Agent Card interoperability | Covered | `ur a2a card`, `/a2a-card`, and opt-in `ur a2a serve` with task lifecycle routes and bearer/delegation auth | Track A2A spec changes and keep the server disabled unless explicitly started |
+| Durable workflows and checkpoints | Covered | resume, rewind, `ur bg` background runs, optional worktrees/PRs, cron/workflow internals, file restore | Publish a checkpointed workflow format for repeated automations |
 | Multi-agent orchestration | Covered | built-in planning, exploration, verification, and general-purpose agents; custom agents | Document reusable team patterns and role selection |
-| Long-term memory | Partial | `/remember`, `/forget`, `.ur/memory`, research notes, team memory, consolidation | Add optional local semantic/vector memory with retention controls |
-| Semantic codebase retrieval | Covered | local embedding-based code index (`ur code-index`), opt-in `CodeSearch` tool, incremental re-index, Ollama embeddings | Add a file-watcher for auto-reindex on change |
+| Long-term memory | Covered | `/remember`, `/forget`, `.ur/memory`, semantic memory, research notes, team memory, consolidation, `ur memory retention` | Tune retention defaults from real long-run telemetry |
+| Semantic codebase retrieval | Covered | local embedding-based code index (`ur code-index`), opt-in `CodeSearch` tool, incremental re-index, auto-reindex watcher, Ollama embeddings | Add richer symbol-aware ranking |
 | AGENTS.md interoperability | Covered | `AGENTS.md` loaded as runtime project context (before `UR.md`), plus imported at `ur init` | Keep aligned as the AGENTS.md spec evolves |
 | Browser and computer-use workflows | Covered | `/browser`, `/chrome`, Playwright-aware tasks, WebSearch, WebFetch, risky-action approval | Add more release fixtures with screenshots and replay assertions |
 | Provenance and citations | Partial | WebFetch source URLs, `/cite`, `/graph`, `/trace`, evidence ledgers | Add claim-to-source mapping for web/MCP answers |
-| Evals and observability | Partial | verifier gates, `.ur/verify.json`, `/verify`, `/trace`, OpenTelemetry hooks, release checks | Publish replayable eval suites and dashboards |
+| Evals and observability | Covered | verifier gates, `.ur/verify.json`, `/verify`, `/trace`, OpenTelemetry hooks, release checks, eval dashboard, benchmark adapters | Publish benchmark numbers from local reproducible suites |
 | Security and prompt-injection resistance | Covered | allow/ask/deny permissions, shell safety analysis, secret scan, untrusted web-content guidance, OS-level execution sandbox (macOS Seatbelt, Linux bubblewrap) | Continuously test web/MCP injection cases |
-| Agent identity and delegated authorization | Partial | MCP OAuth/XAA helpers, local trust boundaries, permission rules | Add portable cross-agent identity only with an opt-in A2A task adapter |
+| Agent identity and delegated authorization | Covered | MCP OAuth/XAA helpers, A2A bearer/delegation tokens, local trust boundaries, permission rules | Keep delegated scopes narrow and auditable |
 | Multimodal workflows | Partial | `/image`, `/video`, `/youtube`, `/voice`, browser workflows | Add model-aware multimodal capability reporting for local Ollama setups |
 | Spec-driven development | Covered | `ur spec` scaffolds requirements/design/tasks under `.ur/specs/`, tracks phase/approvals, and runs the Spec Kit / Kiro task list one task at a time | Add bidirectional sync with an external `specs/` directory |
 | Capability-aware model escalation | Covered | `ur escalate` selects fast/oracle tiers from `model-doctor`, runs routine work fast, and auto-escalates hard/failed work to the strong local model | Learn per-model success rates to tune the difficulty threshold |
 | Best-of-N agent judging | Covered | `ur arena` runs N agents per task in isolated worktrees and judges diffs with the self-review gate; winner is selectable/appliable | Add an optional model judge alongside the deterministic scorer |
 | Self-healing CI | Covered | `ur ci-loop` runs a command, summarizes failures, invokes a fix agent, and re-runs with bounded retries; commits/pushes are self-review gated | Wire to `ur trigger` so a failed CI webhook auto-launches the loop |
-| Verifiable artifacts | Covered | `ur artifacts` records plans/diffs/test-runs with approve/reject/feedback under `.ur/artifacts/` | Attach browser-QA screenshots and link artifacts to claim-ledger entries |
+| Verifiable artifacts | Covered | `ur artifacts` records plans/diffs/test-runs with approve/reject/feedback under `.ur/artifacts/`; comments steer active background agents through stream-json inbox injection | Attach browser-QA screenshots and link artifacts to claim-ledger entries |
+| Native IDE review | Covered | `ur ide diff` writes `.ur/ide/diffs/` bundles and `extensions/vscode-ur-inline-diffs/` provides a VS Code tree/webview/comment surface | Add JetBrains packaging if demand appears |
 
 ## v1.13.9 Direct CLI Surfaces
 
@@ -83,13 +90,11 @@ ur artifacts capture-tests --command "bun test"
 
 ## A2A Position
 
-UR now exports an Agent Card so other tools can discover what UR is and what it
-can do. That is intentionally different from running a network-facing task
-server. A full A2A task endpoint should be opt-in because it changes UR from a
-local CLI into a service that accepts remote work.
-
-The current card is discovery metadata. It describes UR's local-first boundary,
-supported skill areas, input/output modes, and provider metadata.
+UR exports an Agent Card so other tools can discover what UR is and what it can
+do. `ur a2a serve` is the opt-in task server: it exposes health, Agent Card,
+task submission, task listing, task status, output, and cancellation routes. It
+refuses off-loopback binds unless bearer or delegation authentication is
+configured, and task execution stays backed by local UR background tasks.
 
 ## Model Runtime Position
 

package/docs/AGENT_UPGRADE_1.15.0.md

@@ -0,0 +1,129 @@
+# UR Agent 1.15.0 Upgrade Notes
+
+UR 1.15.0 keeps the agent local-first and Ollama-centered. The new agent
+platform surfaces do not add Anthropic, OpenAI, or other direct provider API
+dependencies. Network-facing behavior remains opt-in.
+
+## New Local Agent Surfaces
+
+### Background agents
+
+```sh
+ur bg run "fix the flaky parser test" --worktree
+ur bg fanout "try three parser fixes" --agents 3 --worktree
+ur bg list
+ur bg status <id>
+ur bg logs <id> --tail 100
+ur bg attach <id>
+ur bg kill <id>
+```
+
+State is stored under `.ur/background/`. Worktrees are stored under
+`.ur/worktrees/`. Pull requests are created only when `--pr` is passed and then
+use the existing local `gh` CLI path.
+
+### Context and memory retention
+
+```sh
+ur config set compaction.autoThreshold 80
+ur memory retention show
+ur memory retention set --ttl-days 90 --max-entries 5000 --decay-days 30
+ur memory retention prune
+```
+
+`compaction.autoThreshold` is a global percentage threshold from 50 to 95. The
+retention command prunes project-local `.ur/memory/*.jsonl` files.
+
+### Code-index watcher
+
+```sh
+ur config set codeIndex.autoReindex true
+ur code-index watch --graph
+ur code-index watch --dry-run --json
+```
+
+The watcher uses local filesystem events and the same local Ollama embedding
+path as `ur code-index build`.
+
+### Artifact steering
+
+```sh
+ur artifacts add --kind plan --title "Plan" --task <bg_id>
+ur artifacts comment <artifact_id> --feedback "Prefer the simpler parser path"
+```
+
+Comments are written to the artifact and to the linked background task inbox so
+long-running work has durable steering feedback. Background workers now run in
+stream-json mode and inject new inbox comments into the child agent as
+`priority: "now"` user turns while the process is active.
+
+### A2A task server
+
+```sh
+ur a2a card
+ur a2a serve --host 127.0.0.1 --port 8765 --token "$UR_A2A_TOKEN"
+ur a2a token mint --secret "$UR_A2A_DELEGATION_SECRET" --scope coding-agent
+```
+
+The server is opt-in. It refuses off-loopback binds unless a static bearer token
+or delegation secret is configured. Task execution is backed by UR background
+tasks and local `ur -p`; no external model provider API is introduced.
+
+Useful routes:
+
+- `GET /healthz`
+- `GET /.well-known/agent-card.json`
+- `POST /a2a/tasks`
+- `GET /a2a/tasks`
+- `GET /a2a/tasks/:id`
+- `GET /a2a/tasks/:id/output`
+- `POST /a2a/tasks/:id/cancel`
+- `DELETE /a2a/tasks/:id`
+
+### IDE inline diff bundles
+
+```sh
+ur ide diff capture --title "Parser fix"
+ur ide diff list
+ur ide diff show diff-1
+ur ide diff comment diff-1 --feedback "Inline note" --file src/parser.ts --line 42
+ur ide diff schema
+```
+
+Bundles are stored under `.ur/ide/diffs/` as a manifest, per-diff metadata, and
+unified patch files. The repo also ships a native VS Code extension at
+`extensions/vscode-ur-inline-diffs/` that lists bundles, opens patch previews,
+and writes comments back into the UR metadata.
+
+### Benchmark adapters
+
+```sh
+ur eval bench list
+ur eval bench swe-bench --file swe.jsonl --name local-swe
+ur eval bench terminal-bench --file terminal.jsonl --name local-terminal
+ur eval bench aider-polyglot --file aider.jsonl --name local-polyglot
+ur eval run local-swe --dry-run
+```
+
+Adapters import local JSON or JSONL exports into UR eval suites. They do not
+download datasets or call external services.
+
+## Release Verification
+
+Run these before publishing:
+
+```sh
+bun test test/agentFeatureCommands.test.ts test/agentDelegation.test.ts test/codeIndex.test.ts
+bun run typecheck
+npm pack --dry-run
+```
+
+Optional local smoke checks:
+
+```sh
+bun src/entrypoints/cli.tsx bg list
+bun src/entrypoints/cli.tsx memory retention show
+bun src/entrypoints/cli.tsx code-index watch --dry-run
+bun src/entrypoints/cli.tsx eval bench list
+bun src/entrypoints/cli.tsx ide diff schema
+```