ur-agent 1.28.0 → 1.29.1

package/CHANGELOG.md

@@ -1,5 +1,33 @@
 # Changelog
 
+## 1.29.1
+
+- Replace fabricated Claude Code and Gemini CLI static model names with
+  provider-scoped CLI model aliases/names that the official CLIs can receive.
+- Reject stale subscription CLI selections such as `claude-code/sonnet-5` before
+  runtime dispatch instead of forwarding `sonnet-5` to Claude Code.
+- Summarize subscription CLI model/account failures with provider, model,
+  suggested action, and an explicit no-cross-provider-fallback note.
+
+## 1.29.0
+
+- Customer release consolidating the multi-provider selection and runtime
+  dispatch work (1.27.5–1.28.1) into a single production line.
+- System-prompt identity now reflects the selected provider and runtime backend
+  (e.g. "running through the Codex CLI provider"), so the assistant's self-report
+  matches the `/model` choice instead of a generic default.
+- Refresh documentation to cover the full provider/model feature set: provider-
+  first `/model` flow, provider-scoped discovery, real subscription CLI dispatch,
+  native API wire formats, runtime backends, and troubleshooting.
+
+## 1.28.1
+
+- Keep the status bar synchronized with in-session provider/model changes from
+  `/model`, `/model <model>`, and `/provider`, instead of waiting for persisted
+  settings to reload.
+- Add a regression test proving the status bar prefers the active in-session
+  provider/model over stale persisted settings.
+
 ## 1.28.0
 
 - Subscription CLI providers (Codex, Claude Code, Gemini, Antigravity) now

package/README.md

@@ -194,13 +194,48 @@ Use quotes for shell values with spaces.
 | llama.cpp | local/server | local OpenAI-compatible server |
 | vLLM | local/server | OpenAI-compatible server |
 
-In the interactive app, `/model` shows providers first and then shows only the
-selected provider's models. OpenAI API and Codex CLI, Claude API and Claude
-Code, and Gemini API and Gemini CLI are separate access paths; subscription
-logins and API keys are not interchangeable. The final confirmation includes
-the runtime backend that will receive the next agent request. Use
-`ur provider status` to show the active provider, model, access type, and
-runtime backend.
+#### Provider-first model selection
+
+In the interactive app, `/model` is a two-step, provider-first picker:
+
+1. **Choose a provider.** Every provider is listed with its display name,
+   internal ID, access type (`subscription`, `api`, `local`, `server`),
+   credential type, and a live connection status (`connected`, `missing`,
+   `unavailable`, `unknown`).
+2. **Choose a model.** Only the selected provider's models are shown, labelled
+   by source: `live` (discovered from the endpoint), `cache` (last discovery),
+   or `static` (predefined). Local/server providers (Ollama, LM Studio,
+   llama.cpp, vLLM) and OpenAI-compatible endpoints are discovered live; API and
+   subscription providers use their curated model list.
+
+Model lists never cross providers: OpenAI API vs Codex CLI, Claude API vs Claude
+Code, and Gemini API vs Gemini CLI are separate access paths, and subscription
+logins and API keys are not interchangeable. The provider/model pair is
+validated before it is saved and again before every request; changing provider
+clears an incompatible model. `ur config set provider X` warns and clears an
+incompatible model, and `ur config set model Y` is validated against the active
+provider. The confirmation shows the selected provider, model, model source, and
+the runtime backend that will receive the next request.
+
+#### Runtime dispatch
+
+The selected provider/model drives every agent request — the assistant's own
+identity line in the system prompt reflects it too:
+
+- **Subscription** providers spawn the official CLI (`codex`, `claude`,
+  `gemini`, `agy`) in non-interactive mode with the scoped model and prompt; a
+  non-zero exit or empty output fails clearly (never placeholder text).
+- **API** providers call each service in its native wire format — Anthropic
+  `x-api-key` + `anthropic-version` on `/v1/messages`, OpenAI `Bearer` on
+  `/v1/chat/completions`, Gemini `x-goog-api-key` on `:generateContent`,
+  OpenRouter on its OpenAI-compatible chat endpoint.
+- **Local/server** providers call the configured endpoint (`/v1/chat/completions`
+  for LM Studio/llama.cpp/vLLM; the native API for Ollama).
+
+Ollama is used only when Ollama is selected. There is no silent cross-provider
+fallback: if dispatch fails, UR reports the selected provider, model, and runtime
+backend. Use `ur provider status` (or `ur provider doctor <id>`) to inspect the
+active provider, model, access type, and backend.
 
 Security policy: UR-AGENT never scrapes browser sessions, extracts OAuth
 tokens, bypasses subscription/quota/region/organization restrictions, proxies a
@@ -269,11 +304,13 @@ viewer mode.
 Example:
 
 ```text
-UR-AGENT v1.25.3 | Provider: Ollama | Auth: local | model: qwen3-coder:480b-cloud | mode: ask | branch: main | tasks: idle | Update: 1.25.2 -> 1.25.3 available
+UR-AGENT v1.29.1 | Provider: Codex CLI | Auth: subscription | model: codex/gpt-5.5 | mode: ask | branch: main | tasks: idle
 ```
 
-If a custom status-line hook is configured, UR-AGENT uses that hook output
-instead of the built-in bar.
+The bar reflects the active in-session provider/model immediately after a
+`/model`, `/model <model>`, or `/provider` change — it does not wait for
+persisted settings to reload. If a custom status-line hook is configured,
+UR-AGENT uses that hook output instead of the built-in bar.
 
 ### IDE Integration