@code-yeongyu/senpi 2026.9.11 → 2026.9.12-3

package/docs/compaction.md

@@ -2,12 +2,12 @@
 
 LLMs have limited context windows. When conversations grow too long, senpi uses compaction to summarize older content while preserving recent work. This page covers both auto-compaction and branch summarization.
 
-**Source files** ([pi-mono](https://github.com/earendil-works/pi-mono)):
-- [`packages/coding-agent/src/core/compaction/compaction.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/compaction.ts) - Auto-compaction logic
-- [`packages/coding-agent/src/core/compaction/branch-summarization.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts) - Branch summarization
-- [`packages/coding-agent/src/core/compaction/utils.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/utils.ts) - Shared utilities (file tracking, serialization)
-- [`packages/coding-agent/src/core/session-manager.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/session-manager.ts) - Entry types (`CompactionEntry`, `BranchSummaryEntry`)
-- [`packages/coding-agent/src/core/extensions/types.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/extensions/types.ts) - Extension event types
+**Source files** ([pi](https://github.com/earendil-works/pi)):
+- [`packages/coding-agent/src/core/compaction/compaction.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/compaction.ts) - Auto-compaction logic
+- [`packages/coding-agent/src/core/compaction/branch-summarization.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts) - Branch summarization
+- [`packages/coding-agent/src/core/compaction/utils.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/utils.ts) - Shared utilities (file tracking, serialization)
+- [`packages/coding-agent/src/core/session-manager.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/session-manager.ts) - Entry types (`CompactionEntry`, `BranchSummaryEntry`)
+- [`packages/coding-agent/src/core/extensions/types.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/extensions/types.ts) - Extension event types
 
 For TypeScript definitions in your project, inspect `node_modules/@code-yeongyu/senpi/dist/`.
 
@@ -120,7 +120,7 @@ Never cut at tool results (they must stay with their tool call).
 
 ### CompactionEntry Structure
 
-Defined in [`session-manager.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/session-manager.ts):
+Defined in [`session-manager.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/session-manager.ts):
 
 ```typescript
 interface CompactionEntry<T = unknown> {
@@ -145,7 +145,7 @@ interface CompactionDetails {
 
 Extensions can store any JSON-serializable data in `details`. The default compaction tracks file operations, but custom extension implementations can use their own structure. Generated and extension-provided summaries store their LLM `usage` when available so session totals include summarization work.
 
-See [`prepareCompaction()`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/compaction.ts) and [`compact()`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/compaction.ts) for the implementation. For direct programmatic summarization, `generateSummary()` returns the summary text and `generateSummaryWithUsage()` returns `{ text, usage }`.
+See [`prepareCompaction()`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/compaction.ts) and [`compact()`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/compaction.ts) for the implementation. For direct programmatic summarization, `generateSummary()` returns the summary text and `generateSummaryWithUsage()` returns `{ text, usage }`.
 
 ## Branch Summarization
 
@@ -188,7 +188,7 @@ This means file tracking accumulates across multiple compactions or nested branc
 
 ### BranchSummaryEntry Structure
 
-Defined in [`session-manager.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/session-manager.ts):
+Defined in [`session-manager.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/session-manager.ts):
 
 ```typescript
 interface BranchSummaryEntry<T = unknown> {
@@ -212,7 +212,7 @@ interface BranchSummaryDetails {
 
 Same as compaction, extensions can store custom data in `details`.
 
-See [`collectEntriesForBranchSummary()`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts), [`prepareBranchEntries()`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts), and [`generateBranchSummary()`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts) for the implementation.
+See [`collectEntriesForBranchSummary()`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts), [`prepareBranchEntries()`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts), and [`generateBranchSummary()`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/branch-summarization.ts) for the implementation.
 
 ## Summary Format
 
@@ -256,7 +256,7 @@ path/to/changed.ts
 
 ### Message Serialization
 
-Before summarization, messages are serialized to text via [`serializeConversation()`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/compaction/utils.ts):
+Before summarization, messages are serialized to text via [`serializeConversation()`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/compaction/utils.ts):
 
 ```
 [User]: What they said
@@ -272,7 +272,7 @@ Tool results are truncated to 2000 characters during serialization. Content beyo
 
 ## Custom Summarization via Extensions
 
-Extensions can intercept and customize both compaction and branch summarization. See [`extensions/types.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/extensions/types.ts) for event type definitions.
+Extensions can intercept and customize both compaction and branch summarization. See [`extensions/types.ts`](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/extensions/types.ts) for event type definitions.
 
 ### session_before_compact
 
@@ -288,7 +288,7 @@ pi.on("session_before_compact", async (event, ctx) => {
   // preparation.fileOps - extracted file operations
   // preparation.tokensBefore - context tokens before compaction
   // preparation.firstKeptEntryId - where kept messages start
-  // preparation.settings - compaction settings
+  // preparation.settings - effective settings after applying model overrides
 
   // branchEntries - all entries on current branch (for custom state)
   // reason - "manual" (/compact), "threshold", or "overflow"
@@ -416,3 +416,29 @@ Configure compaction in `~/.senpi/agent/settings.json` or `<project-dir>/.senpi/
 | `keepRecentTokens` | `20000` | Recent tokens to keep (not summarized) |
 
 Disable proactive auto-compaction with `"enabled": false`. You can still compact manually with `/compact`, and a context the provider rejects as too large is still compacted once and retried automatically.
+
+### Per-model overrides
+
+Use `compaction.modelOverrides` to tune token budgets for different models:
+
+```json
+{
+  "compaction": {
+    "reserveTokens": 16384,
+    "keepRecentTokens": 20000,
+    "modelOverrides": {
+      "some-provider/big-model": {
+        "reserveTokens": 400000
+      }
+    }
+  }
+}
+```
+
+For a model with a 1M context window, this override triggers compaction above 600K tokens and keeps the ordinary 20000 recent tokens. Other models retain the ordinary 16384-token reserve. `reserveTokens` also influences summarization output limits, capped by the model's maximum output tokens; it is not solely a trigger threshold.
+
+Keys are exact, case-sensitive `provider/modelId` values, including any slashes within the model ID. Each `reserveTokens` and `keepRecentTokens` value falls back independently from the model override to the ordinary setting to the built-in default. Values must be non-negative safe integers. Invalid values in the matching model override produce an error when read; only omitted fields fall back to the ordinary setting. Model override entries must be objects. Invalid ordinary token settings produce an error when read, even if the active model has a valid override. Only omitted ordinary values use built-in defaults. `enabled` remains global, not model-specific.
+
+These resolved values are used for manual compaction, all automatic threshold checks, overflow recovery, and extension-visible `preparation.settings`. Model switches affect subsequent checks and compactions without changing ordinary settings. Compaction already in progress uses the model and settings captured for that operation. Branch summarization settings are unaffected.
+
+Overrides work in both global and project settings. The files merge recursively before lookup, so a global model-specific value beats a project-wide fallback; a project must override that model entry to change it. See [settings.md](settings.md#per-model-compaction-overrides) for details.

package/docs/containerization.md

@@ -13,6 +13,7 @@ There are two general options. You can either
 | Gondolin extension | Built-in tools and `!` commands | Local micro-VM isolation while keeping auth on host | See [`examples/extensions/gondolin/`](../examples/extensions/gondolin/). |
 | Plain Docker | Whole `senpi` process in a local container | Simple local isolation | Provider API keys enter the container. |
 | OpenShell | Whole `senpi` process in a policy-controlled sandbox | Local or remote managed sandbox | Requires an OpenShell gateway |
+| Docker Sandboxes | Whole `senpi` process in a managed sandbox | Local isolation with provider keys kept on the host | Requires Docker Sandboxes (`sbx`) and a kit that includes `senpi`. |
 
 Extensions run wherever the `senpi` process runs. If you run host `senpi` with a tool-routing extension, other custom extension tools still run on the host unless they also delegate their operations.
 
@@ -109,3 +110,47 @@ openshell sandbox download senpi-sandbox /workspace/repo ./repo-out
 OpenShell providers can keep raw model API keys outside the sandbox.
 When inference routing is configured, code inside the sandbox can call `https://inference.local`, and the gateway injects the configured provider credentials upstream.
 Configure senpi to use the corresponding OpenAI-compatible or Anthropic-compatible endpoint if you want model traffic to use this route.
+
+## Docker Sandboxes
+
+[Docker Sandboxes](https://docs.docker.com/ai/sandboxes/) is a managed sandbox runtime from Docker that runs the whole `senpi` process inside a sandbox.
+It is one of the container boundaries [No Built-in Sandbox](security.md#no-built-in-sandbox) points to.
+
+Unlike the Plain Docker pattern above, the provider credential is not passed into the container.
+The sandbox receives a sentinel value instead, and the `sbx` proxy substitutes the real credential on egress to `api.anthropic.com`.
+Credentials are wired at creation time, so store yours on the host before you create the sandbox.
+
+For a Claude Pro/Max subscription, run `claude setup-token` on a machine with Claude Code, then store the result on the host.
+If an `anthropic` secret is already bound, remove it first: otherwise the proxy adds an `x-api-key` header alongside the Bearer token and Anthropic rejects the request.
+`sbx secret set-custom` reads the token from stdin, so it stays out of shell history.
+
+```bash
+sbx secret rm anthropic
+
+sbx secret set-custom \
+  --host api.anthropic.com \
+  --env ANTHROPIC_OAUTH_TOKEN \
+  --placeholder 'sk-ant-oat01-{rand}'
+```
+
+The sandbox gets an OAuth-shaped placeholder, not the real token, and the proxy swaps it on egress to that host; `ANTHROPIC_OAUTH_TOKEN` is a variable senpi already reads and prefers over an API key, so no extra senpi configuration is needed.
+
+For an API key, store it with `sbx secret set anthropic` instead. The kit wires it the same way, as a sentinel the proxy substitutes on egress.
+
+With the credential stored, launch `senpi` from the project you want mounted. The community `pi-kit` bakes upstream `pi` into its image, not `senpi`, so use a kit that installs `@code-yeongyu/senpi` (or install it inside the sandbox) and start `senpi` from it:
+
+```bash
+sbx run --kit "<your-senpi-kit>" senpi
+```
+
+The current directory becomes the sandbox workspace.
+
+Do not authenticate from inside the sandbox: `/login` there writes a real token into the container and defeats the proxy model.
+
+Scripted use works the same way:
+
+```bash
+sbx exec <sandbox-name> -- senpi -p "list the failing tests"
+```
+
+See the upstream [pi kit documentation](https://github.com/docker/sbx-kits-contrib/tree/main/pi) for the credential matrix, troubleshooting, and pinning; the same secret wiring applies to a `senpi` kit.

package/docs/development.md

@@ -19,6 +19,19 @@ Run from source:
 
 The script can be run from any directory. Senpi keeps the caller's current working directory.
 
+### Experimental remote harness
+
+The upstream remote harness server/client integration lives under `packages/coding-agent/src/experimental/` and is development-only. Run it from the repository with:
+
+```bash
+PI_EXPERIMENTAL=1 ./pi-test.sh server
+PI_EXPERIMENTAL=1 ./pi-test.sh client
+```
+
+`PI_SERVER_DIR` overrides the server profile and socket directory (default: `~/.pi/server`). `PI_SERVER_ID` selects the logical server ID when `--server-id` is omitted.
+
+The `experimental/plugin` subpath and the experimental server/client commands resolve only under the `source` condition in a checkout. Their implementations are excluded from npm packages and standalone binaries. Unlike upstream, `@earendil-works/pi-client` and `@earendil-works/pi-protocol` remain runtime dependencies of `@code-yeongyu/senpi`, and the `./client` entry point stays published; only the experimental sources are source-only. The local SDK and stdio RPC API are unchanged.
+
 ## Forking / Rebranding
 
 This repo is itself a rebrand of upstream `pi-mono` to `senpi`. The runtime identity (CLI name, config dir, env var prefix) is configured via `package.json`:
@@ -69,6 +82,12 @@ bun run --cwd packages/coding-agent test -- test/specific.test.ts
 bun run test --workspace packages/coding-agent -- test/specific.test.ts
 ```
 
+### Published package smoke test
+
+After building, run `npm run check:package-install`. It packs the public packages and installs only coding-agent as a direct dependency in a temporary directory outside the repository. Local tarball overrides select declared transitive dependencies without installing development-only packages. The check verifies SDK imports and CLI startup without credentials or model requests.
+
+`npm run check` also checks runtime dependency declarations and rejects excluded development sources pulled into a package's build through imports.
+
 ## Project Structure
 
 ```

package/docs/docs.json

@@ -93,6 +93,10 @@
 		{
 			"title": "Reference",
 			"items": [
+				{
+					"title": "Environment Variables",
+					"path": "environment-variables.md"
+				},
 				{
 					"title": "Session Format",
 					"path": "session-format.md"

package/docs/environment-variables.md

@@ -105,3 +105,5 @@ These variables are read by Pi itself:
 | `PI_IMAGE_GEN_PROVIDER` | Pin a specific configured gateway provider for image generation (for example `quotio-openai`). The provider must have a resolvable API key and base URL in `models.json`. When unset, the credential resolver picks the best available source automatically |
 
 Provider credentials such as `ANTHROPIC_API_KEY`, `OPENAI_API_KEY`, and cloud-provider configuration are listed in [Providers](providers.md#environment-variables-or-auth-file).
+
+`PI_SERVER_DIR` and `PI_SERVER_ID` apply only to the source-only [experimental remote harness](development.md#experimental-remote-harness), not distributed builds.

package/docs/extensions.md

@@ -1095,6 +1095,12 @@ Access to models, providers, and resolved authentication. `ctx.modelRegistry.get
 
 `ctx.scopedModels` is the read-only list of models scoped to the current session — the same set the `/scoped-models` command shows. It is resolved at session start from the `--models` CLI flag and the `enabledModels` setting (matched against the available catalogue with minimatch on `provider/modelId` or a bare `modelId`). It is empty when no scoping is configured, meaning every available model is usable. Each entry is `{ model, thinkingLevel? }`, where `thinkingLevel` is set only when a pattern pinned it (e.g. `anthropic/*:high`). Use it to populate a model picker that mirrors the built-in one instead of enumerating the whole catalogue via `ctx.modelRegistry.getAvailable()`.
 
+#### Streaming model calls
+
+Use `ctx.modelRegistry.streamSimple(model, context, options)` for provider-neutral options such as `reasoning`, or `stream()` for API-specific options. Both use configured providers and resolve authentication, including for providers registered with `pi.registerProvider()`. Use these instead of `pi-ai/compat` streaming functions, which cannot see extension provider registrations.
+
+Both return an `AssistantMessageEventStream`. Iterate it for response events and await `.result()` for the final message. Setup failures produce error events and error results.
+
 ### ctx.signal
 
 The current agent abort signal, or `undefined` when no agent turn is active.
@@ -1276,7 +1282,7 @@ Options:
 
 ### ctx.navigateTree(targetId, options?)
 
-Navigate to a different point in the session tree:
+Navigate to a different point in the session tree. Rejects with `SessionStreamingError` while an agent response is streaming, even with `summarize: false`; that leaves the active branch unchanged and rejects the promise rather than returning `{ cancelled: true }`. Wait for the response to finish (for example, with `await ctx.waitForIdle()` in a command handler) and retry:
 
 ```typescript
 const result = await ctx.navigateTree("entry-id-456", {
@@ -2443,14 +2449,14 @@ See [examples/extensions/tool-override.ts](../examples/extensions/tool-override.
 **Your implementation must match the exact result shape**, including the `details` type. The UI and session logic depend on these shapes for rendering and state tracking.
 
 Built-in tool implementations:
-- [read.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/read.ts) - `ReadToolDetails`
-- [bash.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/bash.ts) - `BashToolDetails`
-- [powershell.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/powershell.ts) - `PowerShellToolDetails`
-- [edit.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/edit.ts)
-- [write.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/write.ts)
-- [grep.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/grep.ts) - `GrepToolDetails`
-- [find.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/find.ts) - `FindToolDetails`
-- [ls.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/core/tools/ls.ts) - `LsToolDetails`
+- [read.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/read.ts) - `ReadToolDetails`
+- [bash.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/bash.ts) - `BashToolDetails`
+- [powershell.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/powershell.ts) - `PowerShellToolDetails`
+- [edit.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/edit.ts)
+- [write.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/write.ts)
+- [grep.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/grep.ts) - `GrepToolDetails`
+- [find.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/find.ts) - `FindToolDetails`
+- [ls.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/core/tools/ls.ts) - `LsToolDetails`
 
 ### Remote Execution
 
@@ -2581,7 +2587,7 @@ export default function (pi: ExtensionAPI) {
 
 ### Custom Rendering
 
-Tools can provide `renderCall` and `renderResult` for custom TUI display. See [tui.md](tui.md) for the full component API and [tool-execution.ts](https://github.com/earendil-works/pi-mono/blob/main/packages/coding-agent/src/modes/interactive/components/tool-execution.ts) for how tool rows are composed.
+Tools can provide `renderCall` and `renderResult` for custom TUI display. See [tui.md](tui.md) for the full component API and [tool-execution.ts](https://github.com/earendil-works/pi/blob/main/packages/coding-agent/src/modes/interactive/components/tool-execution.ts) for how tool rows are composed.
 
 By default, tool output is wrapped in a `Box` that handles padding and background. A defined `renderCall` or `renderResult` must return a `Component`. If a slot renderer is not defined, `tool-execution.ts` uses fallback rendering for that slot.
 
@@ -2749,6 +2755,10 @@ The DIY lifecycle is:
 - **Anthropic**
   - **Models:** Sonnet, Opus, Fable version 4.5 or newer (without Haiku)
   - **Native representation:** Deferred definitions use `defer_loading`; the load point uses `tool_reference` content.
+- **Fireworks Messages API**
+  - **Native representation:** Deferred definitions use `defer_loading`; the load point uses `tool_reference` content.
+  - **Loader names:** Use `ToolSearch` or `tool_search` for prefix deferral. Other loader names still work, but Fireworks includes the loaded schemas in the initial tool prefix, losing the cache benefit.
+  - This does not change API routing: Fireworks GLM models and Kimi K3 use Chat Completions, not Messages.
 - **OpenAI**
   - **Models:** `gpt-5.4` and newer family
   - **Native representation:** Senpi adds completed client `tool_search_call` and `tool_search_output` items at the load point.
@@ -3193,6 +3203,7 @@ export default function (pi: ExtensionAPI) {
 **Key points:**
 - Extend `CustomEditor` (not base `Editor`) to get app keybindings (escape to abort, ctrl+d, model switching)
 - Call `super.handleInput(data)` for keys you don't handle
+- Custom editors keep the standalone working row by default. Pass `{ embedWorkingStatus: true }` as the fourth `CustomEditor` constructor argument to use the built-in editor-border spinner instead.
 - Factory receives `tui`, `theme`, and `keybindings` from the app
 - Use `ctx.ui.getEditorComponent()` before `setEditorComponent()` to wrap the previously configured custom editor
 - Pass `undefined` to restore default: `ctx.ui.setEditorComponent(undefined)`

package/docs/keybindings.md

@@ -86,10 +86,12 @@ The dedicated history actions always change history entries, regardless of the c
 
 ### TUI Fullscreen Viewport
 
-These actions apply when interactive mode uses `--tui-mode fullscreen` and target the primary transcript scroll region. Two-finger trackpad and mouse-wheel input scroll the region under the pointer, falling back to the transcript over the fixed editor/status/footer dock. Clicking an OSC 8 hyperlink opens it in the default handler. Dragging with the primary mouse button selects text and copies it to the clipboard; holding at the transcript's top or bottom edge auto-scrolls into off-screen content. See [Terminal setup](terminal-setup.md) for terminal-specific mouse and trackpad behavior.
+These actions apply when interactive mode uses `--tui-mode fullscreen` and target the primary transcript scroll region. Two-finger trackpad and mouse-wheel input scroll the region under the pointer, falling back to the transcript over the fixed editor/status/footer dock. Clicking an OSC 8 hyperlink opens it in the default handler. Dragging with the primary mouse button selects text and copies it to the clipboard; holding at the transcript's top or bottom edge auto-scrolls into off-screen content. While the transcript is scrolled up, a clickable "Jump to latest message" label on its bottom row shows the `tui.altScreen.bottom` shortcut. See [Terminal setup](terminal-setup.md) for terminal-specific mouse and trackpad behavior.
 
 Fullscreen transcript bindings take precedence over editor bindings. The default unmodified navigation keys therefore control the transcript in fullscreen mode, while their `ctrl` variants continue to control the editor. Outside fullscreen mode, both variants control the editor.
 
+The transcript search panel shows the configured previous/next shortcuts and clickable arrow controls. Press `tui.altScreen.search` again, or use `tui.altScreen.searchClose`, to close it.
+
 | Key | Default mode | Fullscreen mode |
 |-----|--------------|-----------------|
 | `home`, `end` | Editor | Transcript |
@@ -149,8 +151,10 @@ This routing remains configurable through the ordinary action bindings. For exam
 |--------|---------|-------------|
 | `app.model.select` | `ctrl+l` | Open model selector |
 | `app.model.cycleForward` | `ctrl+p` | Cycle to next model |
-| `app.model.cycleBackward` | `shift+ctrl+p` | Cycle to previous model |
+| `app.model.cycleBackward` | `shift+ctrl+p` (`alt+p` on Windows and WSL) | Cycle to previous model |
+| `app.models.save` | `ctrl+s` | Save the selected default model or scoped model configuration to settings |
 | `app.thinking.cycle` | `shift+tab` | Cycle thinking level |
+| `app.thinking.save` | `ctrl+s` | Save current thinking level to settings |
 | `app.thinking.toggle` | `ctrl+t` | Collapse or expand thinking blocks |
 
 ### Display and Message Queue
@@ -161,6 +165,7 @@ This routing remains configurable through the ordinary action bindings. For exam
 | `app.message.copy` | `ctrl+x` | Copy the selected message in `/tree`; otherwise copy the last assistant message, or the active fullscreen text selection when `fullscreenCopyOnSelect` is disabled |
 | `app.message.followUp` | `alt+enter` | Queue follow-up message |
 | `app.message.dequeue` | `alt+up` | Restore queued messages to editor |
+| `app.question.answer` | `alt+a` | Open the pending async question (Enter on an empty editor and `/answer` do the same without a chord; on macOS the Option-composed glyph of the bound letter is accepted too, except for the dead keys e, i, n, u) |
 
 ### Tree Navigation
 

package/docs/models.md

@@ -586,7 +586,7 @@ For providers with partial OpenAI compatibility, use the `compat` field.
 | `supportsOpenAIGrammarTools` | Whether OpenAI-compatible APIs emit custom Lark/regex grammar tools. When `false`, grammar-constrained tools fall back to normal function tools. Default: `false`; the built-in model catalog enables it for GPT-5+ models on OpenAI, OpenAI Codex, Azure OpenAI, GitHub Copilot, opencode, and Cloudflare AI Gateway. |
 | `supportsRemoteCompactionV2` | For `openai-responses`, opt a verified proxy into native Responses remote compaction v2. Unknown custom proxies default to `false`; official OpenAI defaults to native support. |
 | `deferredToolsMode` | Use provider-specific deferred tool serialization. Currently only `"kimi"` is supported for Kimi's OpenAI-compatible Chat Completions format. |
-| `supportsLongCacheRetention` | Whether the provider accepts long cache retention when cache retention is `long`: `prompt_cache_retention: "24h"` for OpenAI prompt caching, or `cache_control.ttl: "1h"` when `cacheControlFormat` is `anthropic`. Default: `true`. |
+| `supportsLongCacheRetention` | Whether the provider accepts long cache retention when cache retention is `long`: `prompt_cache_options.ttl: "30m"` for GPT-5.6+ Responses models, `prompt_cache_retention: "24h"` for earlier OpenAI models, or `cache_control.ttl: "1h"` when `cacheControlFormat` is `anthropic`. Default: `true`. |
 | `openRouterRouting` | OpenRouter provider routing preferences. This object is sent as-is in the `provider` field of the [OpenRouter API request](https://openrouter.ai/docs/guides/routing/provider-selection). |
 | `vercelGatewayRouting` | Vercel AI Gateway routing config for provider selection (`only`, `order`) |
 

package/docs/providers.md

@@ -272,7 +272,7 @@ senpi
 
 OpenGateway is an OpenAI-compatible multi-provider gateway serving OpenAI, Anthropic, Google, xAI, Moonshot, DeepSeek, ZAI, MiniMax, and Qwen models through one API key. Issue a key at <https://opengateway.ai/api-keys>, then `/login` and select **OpenGateway**, or export `OPENGATEWAY_API_KEY`. The data plane is `https://apis.opengateway.ai`; model ids use the gateway's `owner/model` format (for example `moonshotai/kimi-k3`, `anthropic/claude-fable-5`).
 
-Reference for environment variables and `auth.json` keys: [`const envMap`](https://github.com/earendil-works/pi-mono/blob/main/packages/ai/src/env-api-keys.ts) in [`packages/ai/src/env-api-keys.ts`](https://github.com/earendil-works/pi-mono/blob/main/packages/ai/src/env-api-keys.ts).
+Reference for environment variables and `auth.json` keys: [`const envMap`](https://github.com/earendil-works/pi/blob/main/packages/ai/src/env-api-keys.ts) in [`packages/ai/src/env-api-keys.ts`](https://github.com/earendil-works/pi/blob/main/packages/ai/src/env-api-keys.ts).
 
 #### Auth File
 

package/docs/rpc.md

@@ -222,10 +222,17 @@ and only then constructs its session writer and runtime. A conflicting alias att
 explicitly before opening another writer.
 
 SessionManager writes, switches, forks, new sessions and imports obtain the same grant before writer creation or
-append-side normalization. Acquired paths are conservatively retained for that worker's entire lifetime, including
-superseded paths after a switch. Each worker may reserve at most 64 paths; an exhausted reservation budget fails
-explicitly. Close or an opening deadline requests worker termination, but does not release reservations or worker
-capacity until the actual exit event. A syscall that cannot yet be interrupted can therefore keep an entry
+append-side normalization. A grant is bound to a LIVE writer, not to the worker's lifetime: every snapshot a fully
+open worker publishes names the session files its live session writers still own, and the host releases each granted
+path that list (and the worker's current session path) no longer names. A session replaced by `new_session`,
+`switch_session`, a fork or an import therefore frees its previous file as soon as the replacement is published:
+another `open_session` for that superseded path opens its own worker instead of attaching or failing, and the
+previous opening spelling stops resolving to the replaced owner. Each worker may reserve at most 64 live paths
+at once; a request past that budget first reconciles against the latest snapshot and only then fails with
+`session_reservation_limit`, which is distinct from `session_path_in_use` (another owner holds the path). Only a
+fully open entry reconciles. An entry that is opening, closing or quarantined keeps every acquired path, because a
+worker stuck in a syscall can still be writing a path it can no longer report. Close or an opening deadline requests
+worker termination, but does not release reservations or worker capacity until the actual exit event. A syscall that cannot yet be interrupted can therefore keep an entry
 internally quarantined after the routing handle has closed. `list_sessions` continues to publish `closing`, not a
 new status: existing clients must not mistake a quarantined worker for a live reattach target. Retry the path only
 after that entry disappears.
@@ -296,7 +303,8 @@ In the response `error` field, machine-matchable:
 
 - `unknown_session`
 - `session_closing`
-- `session_path_in_use` (path held by an opening, closing, quarantined, or superseded owner; a fully-open current owner is attached instead)
+- `session_path_in_use` (path held by an opening, closing, or quarantined owner; a fully-open current owner is attached instead, and a path a live owner has superseded is released rather than held)
+- `session_reservation_limit` (this worker already holds 64 live session paths; the open or session replacement was refused without disturbing the existing session)
 - `missing_session_id` (session-scoped command without `sessionId` in multi mode)
 - `multi_session_disabled` (`open_session` in classic mode)
 - `invalid_path` (relative `sessionPath`/`cwd`)
@@ -313,7 +321,7 @@ Strict FIFO per session; one total stdout order; cross-session order unspecified
 
 ### Duplicate/idempotency
 
-Duplicate `open_session` while a path reservation is held by a fully-open session → ATTACH (`attached: true`, same handle); while held by an `opening`/`closing` entry (including internal quarantine) → `session_path_in_use`. `close_session` releases one attachment; the runtime is disposed only when the last attachment closes. A close for an entry already `closing` joins its in-flight teardown. `close_session` on unknown/already-closed → `unknown_session` error. The grace window is configurable by the host through `SENPI_RPC_CLOSE_GRACE_MS`. Request `id`s are client-owned; the server echoes them without dedup.
+Duplicate `open_session` while a path reservation is held by a fully-open session → ATTACH (`attached: true`, same handle); while held by an `opening`/`closing` entry (including internal quarantine) → `session_path_in_use`. A path whose owner has already replaced it with another session file is no longer held: that open allocates a new worker and resumes the file. `close_session` releases one attachment; the runtime is disposed only when the last attachment closes. A close for an entry already `closing` joins its in-flight teardown. `close_session` on unknown/already-closed → `unknown_session` error. The grace window is configurable by the host through `SENPI_RPC_CLOSE_GRACE_MS`. Request `id`s are client-owned; the server echoes them without dedup.
 
 ## Protocol Overview
 
@@ -392,7 +400,7 @@ image payloads are not counted toward this text limit.
 
 #### steer
 
-Queue a steering message while the agent is running. It is delivered after the current assistant turn finishes executing its tool calls, before the next LLM call. Skill commands and prompt templates are expanded. Extension commands are not allowed (use `prompt` instead).
+Queue a steering message while the agent is running. It is delivered after the current assistant turn finishes executing its tool calls, before the next LLM call. Skill commands and prompt templates are expanded. Extension commands are not allowed (use `prompt` instead). Like `prompt`, queued input runs extension `input` handlers with `source: "rpc"`.
 
 ```json
 {"type": "steer", "message": "Stop and do this instead"}
@@ -414,7 +422,7 @@ See [set_steering_mode](#set_steering_mode) for controlling how steering message
 
 #### follow_up
 
-Queue a follow-up message to be processed after the agent finishes. Delivered only when agent has no more tool calls or steering messages. Skill commands and prompt templates are expanded. Extension commands are not allowed (use `prompt` instead).
+Queue a follow-up message to be processed after the agent finishes. Delivered only when agent has no more tool calls or steering messages. Skill commands and prompt templates are expanded. Extension commands are not allowed (use `prompt` instead). Extension `input` handlers run with `source: "rpc"`.
 
 ```json
 {"type": "follow_up", "message": "After you're done, also do this"}

package/docs/sdk.md

@@ -111,6 +111,8 @@ interface AgentSession {
 }
 ```
 
+`session.navigateTree()` rejects with `SessionStreamingError` while an agent response is streaming, even with `summarize: false`. It does not queue navigation or return `{ cancelled: true }` for that conflict. Wait for the response to finish (for example, with `await session.waitForIdle()`) and retry. Rejection leaves the active branch unchanged.
+
 Session replacement APIs such as new-session, resume, fork, and import live on `AgentSessionRuntime`, not on `AgentSession`.
 
 ### createAgentSessionRuntime() and AgentSessionRuntime
@@ -788,6 +790,11 @@ const { session: opened } = await createAgentSession({
   sessionManager: SessionManager.open("/path/to/session.jsonl"),
 });
 
+// Resume a session kept outside the filesystem, e.g. in a database
+const { session: restored } = await createAgentSession({
+  sessionManager: SessionManager.inMemory(process.cwd(), { id: sessionId }, entries),
+});
+
 // List sessions
 const currentProjectSessions = await SessionManager.list(process.cwd());
 const allSessions = await SessionManager.listAll(process.cwd());