snowykr · snowykr · Aug 7, 2026 · Aug 7, 2026 · Aug 7, 2026 · Aug 7, 2026
diff --git a/.gitignore b/.gitignore
@@ -13,6 +13,8 @@ target/
 *.node
 *.node.build.json
 *.b64.js
+# `bun build --compile` leaves these next to the entrypoint when it is interrupted.
+*.bun-build
 
 # Environment
 .env

diff --git a/AGENTS.md b/AGENTS.md
@@ -50,6 +50,16 @@ bun run dev:doctor          # verify PATH resolution of `gjc` points at this wor
 bun run install:defaults    # (re)install bundled default definitions
 ```
 
+Removing build output (never touches sources, `node_modules/`, `.gjc/` state, or `artifacts/` evidence):
+
+```sh
+bun run clean                # dist/, binaries/, coverage/, stray *.bun-build, *.tsbuildinfo
+bun run clean:native         # also drop compiled .node addons (rebuild via build:native)
+bun scripts/clean.ts --dry-run  # list targets without deleting
+```
+
+`clean` removes `packages/coding-agent/dist/`, so a `--binary`-linked `gjc` (see `dev:doctor`) stops resolving until you run `bun run --cwd=packages/coding-agent build` again. Source-linked setups are unaffected.
+
 Verification (never run `tsc`/`npx tsc` directly at repo root; use these):
 
 ```sh

diff --git a/docs/environment-variables.md b/docs/environment-variables.md
@@ -64,6 +64,7 @@ These are consumed via `getEnvApiKey()` (`packages/ai/src/stream.ts`) unless not
 | `OPENROUTER_API_KEY`            | OpenRouter auth                                  | Using OpenRouter models                                        | Also used by image tool when preferred/auto provider is OpenRouter                                  |
 | `MISTRAL_API_KEY`               | Mistral auth                                     | Using Mistral models                                           |                                                                                                     |
 | `ZAI_API_KEY`                   | z.ai auth                                        | Using z.ai models                                              | Also used by z.ai web search provider                                                               |
+| `JUNIE_API_KEY`                 | JetBrains AI (Junie) auth                        | Using `jetbrains-junie` models                                 | Access token from [junie.jetbrains.com/cli](https://junie.jetbrains.com/cli); sent as `Authorization: Bearer`  |
 | `MINIMAX_API_KEY`               | MiniMax auth                                     | Using `minimax` provider                                       |                                                                                                     |
 | `AZURE_OPENAI_API_KEY`          | Azure OpenAI auth                               | Using `azure-openai` / `azure-openai-responses` models         | Pair with `AZURE_OPENAI_BASE_URL` or `AZURE_OPENAI_RESOURCE_NAME`                                   |
 | `MINIMAX_CODE_API_KEY`          | MiniMax Code auth                                | Using `minimax-code` provider                                  |                                                                                                     |

diff --git a/docs/external-control-readiness.md b/docs/external-control-readiness.md
@@ -90,6 +90,8 @@ See [Environment Variables](./environment-variables.md#11-acp-permission-handlin
 
 GJC's ACP session configuration carries the spec-defined `category` on the Mode, Model, and Thinking select options (`mode`, `model`, `thought_level`), which lets ACP clients such as Paseo discover models and thinking levels without provider-specific metadata. The model catalog is filtered to providers with usable stored credentials (`providers.list/active`), falling back to the full catalog on session hosts that do not expose that query.
 
+Model profiles also appear in the ordinary **Model** picker as synthetic entries under the reserved namespace, e.g. `gajae-code/codex-eco` (displayed with the profile label, such as "Codex Eco"). Selecting one through the ACP `Model` select immediately switches the live session to the full profile without persisting `modelProfile.default`; persistence remains an explicit `/model` TUI choice or `gjc --mpreset codex-eco --default`. Only profiles whose providers have usable stored credentials are selectable; synthetic rows are already availability-filtered by the session host, so the Q29 active-provider filter never drops them. An unavailable-but-active profile stays visible as the current readback and, if selected, fails with the existing authentication-required error. The separate ACP startup `--mpreset`/Q27 `Preset` select is likewise session-scoped and non-persistent.
+
 Sessions launched through an ACP client (e.g. `paseo run --provider gjc/...`) are broker-managed and appear in ACP `session/list`, so Paseo's import flow can attach them. Interactive `gjc` sessions host their own SDK endpoint and are not broker-registered, so they are not listed by ACP clients; use the GJC SDK/notifications surface to control those sessions.
 
 ## ACP conformance and Air release gates

diff --git a/docs/models.md b/docs/models.md
@@ -251,6 +251,12 @@ gjc --mpreset opencodego --default
 ```
 
 The `/model` command opens to a preset landing view: presets are grouped by provider with live auth marks (✓/✗), highlighting a group expands its tiers, and selecting a tier shows the full role→model preview before applying for the session or as default. Typing jumps straight to model search, and `Browse all models` opens the classic tabbed model selector. In `/login`, `Add custom provider` is the first option for configuring credentials needed by custom or profile-required providers; after a successful provider login, the matching preset is recommended automatically.
+External SDK/ACP clients (e.g. the Paseo TUI) can select profiles like ordinary
+models: the SDK `models.list/current` (Q10) catalog exposes every usable profile
+as a synthetic `gajae-code/<profile>` entry (e.g. `gajae-code/codex-eco`), and
+selecting one through `model.set` (or the ACP Model picker) activates the
+profile for the live session only. Persisting a profile remains an explicit TUI
+choice, mirroring `gjc --mpreset <name> --default`. See [SDK model profiles](./sdk.md#model-profiles-as-synthetic-models-gajae-codeprofile).
 
 MiniMax's OpenAI-compatible endpoint rejects multiple system messages and emits thinking in `reasoning_content`, so pin the public-safe compatibility fields when hand-authoring a custom provider:
 
@@ -283,6 +289,35 @@ providers:
     models:
       - id: glm-4.6
 ```
+
+### JetBrains AI (Junie)
+
+`jetbrains-junie` is a first-class provider serving JetBrains-hosted models through the documented
+Ingrazzio gateway (`https://ingrazzio-cloud-prod.labs.jb.gg`).
+
+Authenticate with an access token generated at [junie.jetbrains.com/cli](https://junie.jetbrains.com/cli):
+
+```sh
+export JUNIE_API_KEY=...
+```
+
+The token is sent as `Authorization: Bearer` — JetBrains AI rejects requests that also carry `x-api-key`, so
+this provider never lets the Anthropic SDK attach one. Usage is billed against your JetBrains AI
+subscription, so bundled per-token costs are zero. There is no OAuth login flow; the environment variable is
+the only supported credential source.
+
+The gateway multiplexes transports by model family:
+
+| Family | Models | Transport | Prompt limit |
+| --- | --- | --- | --- |
+| Claude | `claude-sonnet-4-6` (default), `claude-sonnet-5`, `claude-opus-4-6`, `claude-opus-4-7`, `claude-opus-4-8`, `claude-opus-5`, `claude-fable-5` | `anthropic-messages` | 1M |
+| GPT | `gpt-5-2025-08-07`, `gpt-5.2-2025-12-11`, `gpt-5.4`, `gpt-5.5`, `gpt-5.6-luna`, `gpt-5.6-sol`, `gpt-5.6-terra` | `openai-completions` | 922K |
+| GPT (Responses-only) | `gpt-5.3-codex` | `openai-responses` | 272K |
+
+All models cap output at 128K. Junie also exposes Gemini and Grok, but those ride a proprietary Grazie
+translation protocol that GJC does not implement, so they are deliberately not bundled. The bare
+`opus`/`sonnet`/`gpt`/`grok` aliases are Junie CLI shorthands the gateway itself rejects.
+
 ### Allowed auth/discovery values
 
 - `auth`: `apiKey` (default), `none`, or `oauth`; for `models.yml` custom models, `oauth` is accepted by schema but does not waive the `apiKey` requirement

diff --git a/docs/sdk-app-guide.md b/docs/sdk-app-guide.md
@@ -182,6 +182,13 @@ Beyond frames, the WS surface exposes typed **control operations**
 `usage.get`, `models.list/current`, `workflow.gates.list`, …). See the
 [SDK wire protocol & machine interfaces](./sdk.md) for the complete catalog.
 
+The `models.list/current` (Q10) catalog also lists model profiles as synthetic
+`gajae-code/<profile>` entries (e.g. `gajae-code/codex-eco`). Treat them as
+logical selections, not API providers: sending the id back through `model.set`
+activates the profile for the live session only. Persisting remains an explicit
+TUI choice. Request Q27 (`models.profiles.list`) when you need the
+full profile catalog including unavailable profiles and their `available`
+status. See [Model profiles as synthetic models](./sdk.md#model-profiles-as-synthetic-models-gajae-codeprofile).
 
 ## Creating and supervising sessions
 

diff --git a/docs/sdk.md b/docs/sdk.md
@@ -283,6 +283,65 @@ as a `model.set` input.
 Malformed reasoning descriptors are not client-recoverable catalog data. The
 query returns the SDK's safe `internal` error rather than exposing a partially
 formed row or descriptor details.
+### Model profiles as synthetic models (`gajae-code/<profile>`)
+
+The Q10 catalog also exposes model profiles as logical synthetic models under
+the reserved provider namespace `gajae-code`, e.g. `gajae-code/codex-eco`.
+These rows let clients (such as ACP model pickers) offer presets like ordinary
+models without provider-specific metadata:
+
+```json
+{
+  "provider": "gajae-code",
+  "id": "codex-eco",
+  "name": "Codex Eco",
+  "contextWindow": 222222,
+  "maxTokens": 8888,
+  "reasoning": false,
+  "thinking": { "validLevels": ["off"] },
+  "current": false
+}
+```
+
+- `gajae-code/<profile>` is a **logical namespace, not a callable provider**. No
+  API transport, credentials, or streaming route is registered for it; send the
+  value back through the generic `model.set` control (or the ACP `Model`
+  select) to activate the profile.
+- Synthetic rows are **availability-filtered**: only profiles whose required and
+  alternative providers have usable stored credentials are listed. The profile
+  id suffix is parsed losslessly after the first namespace slash, so profile ids
+  containing additional slashes or punctuation round-trip exactly.
+- `contextWindow`/`maxTokens` mirror the profile's resolvable default model when
+  available and otherwise fall back to the shared unknown-model constants
+  (222222 / 8888); the profile's real default model remains authoritative.
+- Synthetic rows are non-reasoning with `validLevels: ["off"]`: a `model.set`
+  on a synthetic id with any thinking level other than `off` is rejected with
+  `invalid_input`, and only an absent or `off` level is forwarded as a session
+  override.
+- **Current-state semantics:** while a profile is active for the session, exactly
+  the synthetic row carries `current: true` with `currentThinkingLevel:
+  "inherit"`, and the underlying concrete row is not marked current. A persisted
+  `modelProfile.default` alone (without an in-session active marker) never
+  creates a synthetic current row. Selecting a concrete `provider/model` clears
+  the active marker and restores concrete current semantics.
+- **Selecting a synthetic profile is session-scoped.** `model.set` with
+  `gajae-code/<profile>` activates the full profile in the live session without
+  writing `modelProfile.default`, `modelRoles`, or
+  `task.agentModelOverrides`. Persisting a profile remains an explicit TUI
+  choice (`/model` → default), mirroring `gjc --mpreset <profile> --default`.
+  Unknown or ambiguous synthetic ids fail with `invalid_input`; missing profile
+  credentials fail with the existing authentication-required error.
+- `gajae-code` is **reserved**: a user-defined `models.yml` provider of the same
+  name disables the synthetic facade (rows are omitted and synthetic selection
+  is rejected) rather than being silently shadowed. Q27 (`models.profiles.list`)
+  remains the full profile catalog with explicit `available` status; Q10 is the
+  availability-aware facade for client selection.
+`config.patch` mutations are serialized through the same session admission
+boundary as profile activation and default-model selection, so a patch racing
+a synthetic `gajae-code/*` selection (or another patch) is applied in a
+deterministic order and is never lost or clobbered by an activation rollback.
+The cost is the same as `model.set`: an external `config.patch` queues behind
+any in-flight prompt admission rather than applying mid-turn.
 
 ## Prompt acceptance, termination, and reconciliation (Q26)
 

diff --git a/package.json b/package.json
@@ -94,9 +94,11 @@
     "stats": "bun --cwd=packages/coding-agent src/cli.ts stats",
     "build": "bun run --workspaces --if-present build",
     "build:native": "bun --cwd=packages/natives run build",
+    "clean": "bun scripts/clean.ts",
+    "clean:native": "bun scripts/clean.ts --native",
     "test": "bun run --parallel test:ts test:rs",
     "test:ts": "bun run test:release && bun run --workspaces --if-present test",
-    "test:release": "bun test scripts/nightly-release.test.ts scripts/release-evidence.test.ts scripts/release-policy.test.ts scripts/release-publish-order.test.ts scripts/restart-sdk-broker.test.ts",
+    "test:release": "bun test scripts/clean.test.ts scripts/nightly-release.test.ts scripts/release-evidence.test.ts scripts/release-policy.test.ts scripts/release-publish-order.test.ts scripts/restart-sdk-broker.test.ts",
     "generate-schemas": "bun scripts/generate-json-schemas.ts",
     "check:schemas": "bun scripts/generate-json-schemas.ts --check",
     "check:public-sync": "bun scripts/check-public-version-sync.ts",

diff --git a/packages/ai/CHANGELOG.md b/packages/ai/CHANGELOG.md
@@ -4,6 +4,7 @@
 ### Added
 
 - Added opt-in `compat.supportsResponsesSessionAffinity` for OpenAI Responses custom relays. When enabled, supported `openai-responses` models may send `session_id` and `x-client-request-id` affinity headers to a custom endpoint; canonical OpenAI routing remains automatic and known non-OpenAI provider IDs remain excluded.
+- Added the `jetbrains-junie` provider, serving JetBrains-hosted models over the documented Ingrazzio gateway `https://ingrazzio-cloud-prod.labs.jb.gg` (#3626). Auth is the officially documented `JUNIE_API_KEY` access token only — no OAuth login flow and no reverse-engineered client credentials. JetBrains AI rejects requests carrying `x-api-key`, so the provider passes `apiKey: null` to the Anthropic SDK and relies solely on the `Authorization: Bearer` header that `buildAnthropicHeaders` already emits for non-Anthropic hosts. The gateway multiplexes transports by family via the `X-LLM-Model` routing header: 7 Claude models on `anthropic-messages` (1M prompt window), 7 GPT models on `openai-completions` and `gpt-5.3-codex` on `openai-responses` (922K and 272K respectively); all cap output at 128K. The GPT lane pins a `/v1`-suffixed base URL because the OpenAI transports append a bare route while the Anthropic one supplies its own prefix. Ids come from Junie CLI's own catalog cross-checked against the 2470.4 jar; Gemini and Grok are excluded because their Grazie translation protocol is not implemented, and the bare `opus`/`sonnet` aliases are CLI shorthands the gateway rejects. Limits are the gateway's probed ceilings, not Junie CLI's smaller per-request budgets.
 
 ### Fixed