ur-agent 1.82.0 → 1.82.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/CHANGELOG.md CHANGED
@@ -1,5 +1,22 @@
1
1
  # Changelog
2
2
 
3
+ ## 1.82.1
4
+
5
+ - Added a mandatory execution/trust kernel to every main, custom, and subagent
6
+ prompt; repository files, tool output, recalled memory, and MCP instructions
7
+ are now explicitly provenance-wrapped as untrusted evidence rather than
8
+ authority. A per-process canary, presentation-boundary redaction, and a final
9
+ post-hook input check prevent hidden-instruction and canary exfiltration.
10
+ - Added provenance-aware authorization for suspicious evidence, regression-safe
11
+ prompt optimization, exact prompt/tool/context/model fingerprints, stronger
12
+ deterministic eval fixtures, hidden behavior checks, and least-privilege
13
+ generated agent specifications with explicit tools, permissions, turn limits,
14
+ and evaluation cases.
15
+ - Added Unsloth as an authenticated, provider-only OpenAI-compatible runtime
16
+ with live model discovery, native streaming, and standard function calls.
17
+ UR never manages Unsloth and forces `enable_tools: false` so all execution
18
+ remains inside UR's permission, sandbox, provenance, and verifier boundary.
19
+
3
20
  ## 1.82.0
4
21
 
5
22
  - Restored the complete UR agent prompt, conversation history, tools, and
package/README.md CHANGED
@@ -284,6 +284,7 @@ ur config set provider openai-api
284
284
  ur config set provider anthropic-api
285
285
  ur config set provider gemini-api
286
286
  ur config set provider openrouter
287
+ ur config set provider unsloth
287
288
  ur config set model qwen2.5-coder:7b
288
289
  ur provider select-model ollama qwen2.5-coder:7b --json
289
290
  ur config set base_url http://localhost:11434
@@ -314,7 +315,7 @@ to the default with `ur config set openai_transport chat-completions`.
314
315
 
315
316
  Provider config accepts canonical IDs and common aliases. Examples:
316
317
  `openai-api`, `anthropic-api`, `gemini-api`, `openrouter`, `ollama`,
317
- `lmstudio`, `LM Studio`, `llama.cpp`, `vllm`, and the subscription CLIs
318
+ `lmstudio`, `LM Studio`, `llama.cpp`, `vllm`, `unsloth` (`Unsloth Studio`), and the subscription CLIs
318
319
  `codex-cli` (`chatgpt`), `claude-code-cli` (`claude`), `gemini-cli` (`gemini`),
319
320
  and `antigravity-cli` (`agy`). Use quotes for shell values with spaces.
320
321
 
@@ -337,6 +338,7 @@ ur connect logout openai-api # clear a stored key
337
338
  | LM Studio | local/server | UR-native | local OpenAI-compatible server |
338
339
  | llama.cpp | local/server | UR-native | local OpenAI-compatible server |
339
340
  | vLLM | local/server | UR-native | OpenAI-compatible server |
341
+ | Unsloth | local/server | UR-native | authenticated user-run Unsloth Studio endpoint (`UNSLOTH_API_KEY`) |
340
342
  | Codex CLI | subscription | external app bridge | official Codex CLI login (`ur auth chatgpt`) |
341
343
  | Claude Code | subscription | external app bridge | official Claude Code login (`ur auth claude`) |
342
344
  | Gemini CLI | subscription | external app bridge | official Gemini Code Assist login (`ur auth gemini`) |
@@ -353,7 +355,7 @@ In the interactive app, `/model` is a two-step, provider-first picker:
353
355
  2. **Choose a model.** Only the selected provider's models are shown, labelled
354
356
  by source: `live` (discovered from the endpoint), `cache` (last discovery),
355
357
  or `static` (predefined). Local/server providers (Ollama, LM Studio,
356
- llama.cpp, vLLM) and OpenAI-compatible endpoints are discovered live; API
358
+ llama.cpp, vLLM, Unsloth) and OpenAI-compatible endpoints are discovered live; API
357
359
  providers use live discovery from their `/models` endpoint once a key is
358
360
  connected (with a curated fallback list before that). Subscription CLIs show
359
361
  their curated model list because the official CLIs expose no models API. The
@@ -394,7 +396,10 @@ identity line in the system prompt reflects it too:
394
396
  selected, Gemini `x-goog-api-key` on `:generateContent`, and OpenRouter on its
395
397
  OpenAI-compatible chat endpoint.
396
398
  - **Local/server** providers call the configured endpoint (`/v1/chat/completions`
397
- for LM Studio/llama.cpp/vLLM; the native API for Ollama).
399
+ for LM Studio/llama.cpp/vLLM/Unsloth; the native API for Ollama). Unsloth is
400
+ provider-only: UR never starts, installs, updates, trains, or loads models in
401
+ Unsloth. UR sends `enable_tools: false` so only UR's permissioned native tools
402
+ can execute.
398
403
  - **Subscription CLIs** (Codex, Claude Code, Gemini, Antigravity) are external
399
404
  app bridges: selecting one dispatches the turn through the vendor's official
400
405
  CLI using your subscription. Log in with `ur auth <provider>`. UR-native
package/RELEASE.md CHANGED
@@ -72,7 +72,8 @@ Version bump checklist (all versioned release surfaces must move together):
72
72
  5. `extensions/jetbrains-ur/build.gradle.kts` `version`
73
73
  6. The `MACRO.VERSION` fallbacks in `src/commands/agent-ci/agent-ci.ts`,
74
74
  `src/services/agents/agenticCi.ts`, and
75
- `src/services/agents/featureScaffolds.ts`
75
+ `src/services/agents/featureScaffolds.ts`, and
76
+ `src/services/agents/trends.ts`
76
77
  7. Add a `CHANGELOG.md` entry, then run `bun run build` so `dist/cli.js`
77
78
  embeds the new version (`bun run release:check` verifies all of this).
78
79