llm-cli-gateway 1.17.3 → 1.17.5

package/dist/flight-recorder.js

@@ -1,30 +1,8 @@
-/**
- * Flight recorder: SQLite-backed request log.
- *
- * Read access for cache-stats / MCP resources / doctor goes through the
- * `queryRequests<T>(sql, ...params)` method exposed on both `FlightRecorder`
- * and `NoopFlightRecorder` (the `FlightRecorderQuery` interface, see bottom
- * of file). This is Option A from
- * docs/plans/cache-awareness.dag.toml#expose-flight-recorder-read-access —
- * a single read-only query surface on the existing class, NOT a sibling
- * read-only SQLite connection. better-sqlite3 in WAL mode handles
- * concurrent readers inside a single process safely, so the additional
- * connection isn't needed and would have to be threaded through
- * GatewayServerRuntime as a separate field.
- *
- * Callers MUST pass parameterised SQL — string-interpolation of untrusted
- * values is unsafe even on a "read-only" query.
- */
 import { chmodSync, existsSync, mkdirSync } from "fs";
 import os from "os";
 import path from "path";
 import { createRequire } from "module";
 const MAX_THINKING_BYTES = 1_000_000;
-/**
- * Idempotent migration: add `cache_read_tokens` / `cache_creation_tokens`
- * columns to the `requests` table if a pre-U23 logs.db is opened. Existing
- * rows keep NULL for the new columns; that is intentional.
- */
 function ensureRequestsCacheColumns(db) {
     const rows = db.prepare("PRAGMA table_info(requests)").all?.() ?? [];
     const names = new Set(rows.map((row) => (row && typeof row.name === "string" ? row.name : "")));
@@ -35,15 +13,6 @@ function ensureRequestsCacheColumns(db) {
         db.exec("ALTER TABLE requests ADD COLUMN cache_creation_tokens INTEGER");
     }
 }
-/**
- * Idempotent v3 migration: add `stable_prefix_hash` / `stable_prefix_tokens`
- * columns plus their index. Populated only for new rows that carry a
- * promptParts structure (slice 1); legacy rows keep NULL forever.
- *
- * Read access for cache-stats / MCP resources / doctor goes through the
- * read-only `queryRequests()` method on FlightRecorder (no separate read
- * connection — better-sqlite3 in WAL mode handles concurrent readers).
- */
 function ensureStablePrefixColumns(db) {
     const rows = db.prepare("PRAGMA table_info(requests)").all?.() ?? [];
     const names = new Set(rows.map((row) => (row && typeof row.name === "string" ? row.name : "")));
@@ -55,13 +24,6 @@ function ensureStablePrefixColumns(db) {
     }
     db.exec("CREATE INDEX IF NOT EXISTS idx_requests_stable_hash ON requests(stable_prefix_hash)");
 }
-/**
- * Idempotent v4 migration (slice κ): add `cache_control_blocks` column
- * to the `requests` table. Counts the caller-supplied content blocks
- * the gateway emitted with an explicit Anthropic `cache_control`
- * marker. Pre-κ rows keep NULL; only κ-opt-in callers ever set the
- * column to a non-NULL integer.
- */
 function ensureCacheControlBlocksColumn(db) {
     const rows = db.prepare("PRAGMA table_info(requests)").all?.() ?? [];
     const names = new Set(rows.map((row) => (row && typeof row.name === "string" ? row.name : "")));
@@ -92,10 +54,8 @@ function truncateThinkingBlocks(blocks) {
             used += bytes;
             continue;
         }
-        // Reserve space for the suffix so total stays within budget
         const budget = Math.max(0, MAX_THINKING_BYTES - used - TRUNCATION_SUFFIX_BYTES);
         if (budget > 0) {
-            // Truncate on code point boundaries by using string iteration
             let charBytes = 0;
             let safeEnd = 0;
             for (const char of block) {
@@ -103,7 +63,7 @@ function truncateThinkingBlocks(blocks) {
                 if (charBytes + charSize > budget)
                     break;
                 charBytes += charSize;
-                safeEnd += char.length; // char.length handles surrogate pairs
+                safeEnd += char.length;
             }
             const sliced = block.slice(0, safeEnd);
             result.push(sliced ? `${sliced}${TRUNCATION_SUFFIX}` : TRUNCATION_SUFFIX);
@@ -174,26 +134,14 @@ export class FlightRecorder {
         this.db
             .prepare("INSERT OR IGNORE INTO _migrations(version, applied_at) VALUES(1, ?)")
             .run(new Date().toISOString());
-        // Migration v2: cache_read_tokens / cache_creation_tokens columns on
-        // pre-U23 logs.db files. ALTER TABLE ADD COLUMN is idempotent only via
-        // a prior PRAGMA table_info() check; better-sqlite3 has no native
-        // "IF NOT EXISTS" for ADD COLUMN.
         ensureRequestsCacheColumns(this.db);
         this.db
             .prepare("INSERT OR IGNORE INTO _migrations(version, applied_at) VALUES(2, ?)")
             .run(new Date().toISOString());
-        // Migration v3: stable_prefix_hash / stable_prefix_tokens columns plus
-        // their index. Populated only for new rows whose request carried a
-        // promptParts structure (slice 1 of cache-awareness); legacy rows keep
-        // NULL intentionally.
         ensureStablePrefixColumns(this.db);
         this.db
             .prepare("INSERT OR IGNORE INTO _migrations(version, applied_at) VALUES(3, ?)")
             .run(new Date().toISOString());
-        // Migration v4: cache_control_blocks (slice κ). Pre-κ rows keep NULL;
-        // only κ-opt-in writes populate this. Aggregates in cache-stats /
-        // MCP resources can use this to separate explicit κ hits from
-        // implicit prefix-cache hits.
         ensureCacheControlBlocksColumn(this.db);
         this.db
             .prepare("INSERT OR IGNORE INTO _migrations(version, applied_at) VALUES(4, ?)")
@@ -203,7 +151,6 @@ export class FlightRecorder {
                 chmodSync(dbPath, 0o600);
             }
             catch {
-                // Best effort permissions hardening.
             }
         }
         const insertRequest = this.db.prepare(`
@@ -292,21 +239,6 @@ export class FlightRecorder {
     logComplete(correlationId, result) {
         this.updateCompleteTxn(correlationId, result);
     }
-    /**
-     * Read-only query over the requests + gateway_metadata tables. Used by
-     * cache-stats / MCP resources / doctor without exposing a second SQLite
-     * connection. better-sqlite3 in WAL mode handles concurrent readers
-     * inside a single process safely.
-     *
-     * Safety:
-     * - Caller MUST pass parameterised SQL — direct string interpolation of
-     *   untrusted values is unsafe.
-     * - The compiled statement's `.readonly` flag is checked at runtime;
-     *   anything that can mutate rows (INSERT/UPDATE/DELETE, including the
-     *   `RETURNING` forms that better-sqlite3 surfaces via `.all()`) throws.
-     *   This blocks the writer-disguised-as-reader vector codex-r1/F3
-     *   flagged, even when the caller is internal gateway code.
-     */
     queryRequests(sql, ...params) {
         const stmt = this.db.prepare(sql);
         if (stmt.readonly === false) {
@@ -318,7 +250,6 @@ export class FlightRecorder {
         return stmt.all(...params);
     }
     flush() {
-        // No-op: better-sqlite3 writes synchronously.
     }
     close() {
         this.db.close();

package/dist/gemini-json-parser.d.ts

@@ -1,23 +1,3 @@
-/**
- * Parsers for Gemini CLI `-o json` (single object) and `-o stream-json`
- * (NDJSON event stream) output.
- *
- * `-o json` emits a single JSON object with:
- *   - `response`: string final model output
- *   - `usageMetadata`: { promptTokenCount, candidatesTokenCount,
- *                        cachedContentTokenCount?, totalTokenCount }
- *
- * `-o stream-json` emits one JSON object per line:
- *   - `{ "type": "init", "session_id": "...", "model": "..." }`
- *   - `{ "type": "message", "role": "user", "content": "..." }`
- *   - `{ "type": "message", "role": "assistant", "content": "...", "delta": true }` (repeated)
- *   - `{ "type": "result", "status": "success", "stats": { "input_tokens": N,
- *        "output_tokens": N, "cached": N, ... } }`
- *
- * Both parsers return null when stdout is unparseable. Both populate the same
- * `GeminiJsonParseResult` shape so `extractUsageAndCost` can branch on
- * outputFormat without further dispatch.
- */
 export interface GeminiUsage {
     input_tokens: number;
     output_tokens: number;
@@ -28,9 +8,4 @@ export interface GeminiJsonParseResult {
     response?: string;
 }
 export declare function parseGeminiJson(stdout: string): GeminiJsonParseResult | null;
-/**
- * Parse Gemini `-o stream-json` NDJSON output. Concatenates assistant `delta`
- * message content into `response`, extracts the terminal `result.stats` payload
- * into `usage`. Returns null when stdout contains no parseable JSON line.
- */
 export declare function parseGeminiStreamJson(stdout: string): GeminiJsonParseResult | null;

package/dist/gemini-json-parser.js

@@ -1,23 +1,3 @@
-/**
- * Parsers for Gemini CLI `-o json` (single object) and `-o stream-json`
- * (NDJSON event stream) output.
- *
- * `-o json` emits a single JSON object with:
- *   - `response`: string final model output
- *   - `usageMetadata`: { promptTokenCount, candidatesTokenCount,
- *                        cachedContentTokenCount?, totalTokenCount }
- *
- * `-o stream-json` emits one JSON object per line:
- *   - `{ "type": "init", "session_id": "...", "model": "..." }`
- *   - `{ "type": "message", "role": "user", "content": "..." }`
- *   - `{ "type": "message", "role": "assistant", "content": "...", "delta": true }` (repeated)
- *   - `{ "type": "result", "status": "success", "stats": { "input_tokens": N,
- *        "output_tokens": N, "cached": N, ... } }`
- *
- * Both parsers return null when stdout is unparseable. Both populate the same
- * `GeminiJsonParseResult` shape so `extractUsageAndCost` can branch on
- * outputFormat without further dispatch.
- */
 export function parseGeminiJson(stdout) {
     const trimmed = stdout.trim();
     if (!trimmed) {
@@ -54,11 +34,6 @@ export function parseGeminiJson(stdout) {
     }
     return result;
 }
-/**
- * Parse Gemini `-o stream-json` NDJSON output. Concatenates assistant `delta`
- * message content into `response`, extracts the terminal `result.stats` payload
- * into `usage`. Returns null when stdout contains no parseable JSON line.
- */
 export function parseGeminiStreamJson(stdout) {
     if (!stdout) {
         return null;
@@ -71,9 +46,6 @@ export function parseGeminiStreamJson(stdout) {
         const trimmed = line.trim();
         if (!trimmed)
             continue;
-        // Gemini stream-json lines are individual JSON objects; non-JSON
-        // chatter (warnings, "Ripgrep not available", etc.) is silently
-        // ignored so a stray banner line doesn't poison usage extraction.
         let event;
         try {
             event = JSON.parse(trimmed);

package/dist/health.d.ts

@@ -13,8 +13,5 @@ export interface ProviderRuntimeHealth {
     providers: Record<string, Pick<ProviderRuntimeStatus, "installed" | "version" | "loginStatus" | "loginCheck">>;
     timestamp: string;
 }
-/**
- * Check health status of PostgreSQL.
- */
 export declare function checkHealth(db: DatabaseConnection): Promise<HealthStatus>;
 export declare function checkProviderRuntimeHealth(): ProviderRuntimeHealth;

package/dist/health.js

@@ -1,7 +1,4 @@
 import { listProviderRuntimeStatuses } from "./provider-status.js";
-/**
- * Check health status of PostgreSQL.
- */
 export async function checkHealth(db) {
     const result = await db.healthCheck();
     const health = {

package/dist/index.d.ts

@@ -13,18 +13,9 @@ import { ClaudeMcpConfigResult, ClaudeMcpServerName } from "./claude-mcp-config.
 import { type MistralAgentMode, type ClaudePermissionMode, type CodexSandboxMode, type CodexAskForApproval, type ClaudeEffortLevel } from "./request-helpers.js";
 import { FlightRecorderLike } from "./flight-recorder.js";
 import { type PromptParts } from "./prompt-parts.js";
-/**
- * Slice 3: structured warning entries attached to tool responses.
- * Distinct from review-integrity warnings (which are text-appended to
- * the user-visible response). These are programmatic signals for caller
- * agents to react to.
- */
 export interface WarningEntry {
-    /** Stable machine-readable code, e.g. "cache_ttl_expiring_soon". */
     code: string;
-    /** Optional human-readable message for surfaces that render text. */
     message?: string;
-    /** Code-specific payload — left open for future warning types. */
     ttlRemainingMs?: number;
     [key: string]: unknown;
 }
@@ -44,7 +35,6 @@ type ExtendedToolResponse = {
         missing?: ClaudeMcpServerName[];
     };
     reviewIntegrity?: ReviewIntegrityResult;
-    /** Slice 3: structured warnings (e.g. cache_ttl_expiring_soon). */
     warnings?: WarningEntry[];
 };
 declare const logger: {
@@ -54,40 +44,9 @@ declare const logger: {
     debug: (message: string, ...args: any[]) => void;
 };
 type GatewayLogger = typeof logger;
-/**
- * Phase 4 slice δ — shared Zod fragments for `maxTurns` / `maxPrice`.
- *
- * Both flags reach the upstream CLIs as decimal-formatted argv strings via
- * `String(N)`. `z.number().int().positive()` alone lets values past
- * `Number.MAX_SAFE_INTEGER` through, after which `String(1e21)` emits
- * scientific notation that Grok and Vibe both reject. The bounds below
- * (safe-integer cap + 10000 ceiling for turns; finite + 10000 USD ceiling
- * for price) guarantee a lossless decimal stringification AND a sane
- * upper bound — no plausible single agent loop exceeds 10k turns or 10k USD.
- */
 export declare const MAX_TURNS_SCHEMA: z.ZodNumber;
 export declare const MAX_TOKENS_SCHEMA: z.ZodNumber;
 export declare const MAX_PRICE_SCHEMA: z.ZodNumber;
-/**
- * Slice λ: shared worktree directive for all 10 `*_request` / `*_request_async`
- * tools. `true` creates a fresh worktree under `<repoRoot>/.worktrees/<uuid>`
- * branched from HEAD. `{ name?, ref? }` lets the caller supply a sanitized
- * name and/or git ref (default ref: HEAD).
- *
- * Lifecycle is gateway-owned: the gateway pre-creates the worktree via
- * `git worktree add`, then spawns the child CLI with `cwd: <worktree-path>`.
- * No `-w` / `--worktree` flag is ever emitted to the underlying CLI. When
- * the request carries a sessionId and the session already has a worktree,
- * that worktree is reused. On session_delete or TTL eviction the gateway
- * runs `git worktree remove --force`.
- *
- * Tool response: when a worktree was used, the successful response stdout
- * is prefixed with `[gateway] worktree=<absolute-path>\n` so callers can
- * parse/use the path without a schema change (slice λ §1.d).
- *
- * NOTE: callers should `.gitignore` the `.worktrees/` directory in their
- * repo (the gateway does NOT auto-gitignore — see slice λ spec Q4).
- */
 export declare const WORKTREE_SCHEMA: z.ZodUnion<[z.ZodBoolean, z.ZodObject<{
     name: z.ZodOptional<z.ZodString>;
     ref: z.ZodOptional<z.ZodString>;
@@ -128,65 +87,16 @@ export interface GatewayServerRuntime {
 export declare function resolveGatewayServerRuntime(deps?: GatewayServerDeps, options?: {
     isolateState?: boolean;
 }): GatewayServerRuntime;
-/**
- * Slice λ: shape returned by `resolveWorktreeForRequest`. `cwd` is what
- * the spawn helpers (`executeCli`, `startJobWithDedup`) consume;
- * `worktreePath` is what the tool handler embeds in the response prefix
- * so callers can discover the path.
- */
 export interface ResolvedWorktree {
     cwd?: string;
     worktreePath?: string;
 }
-/**
- * Slice λ: resolve a request's worktree directive into a spawn cwd.
- *
- * - `worktreeOpt` is the Zod-validated input value (boolean |
- *   `{ name?, ref? }` | undefined).
- * - When the request has a session AND the session already has a
- *   `metadata.worktreePath`, that path is reused (resume semantics).
- *   The reused path is returned without touching git; if the directory
- *   was externally removed between requests, the next CLI invocation
- *   will surface the error naturally.
- * - When no reusable worktree exists, `createWorktree` runs; on success
- *   the new path is written to `session.metadata` (only when a session
- *   exists — request-scoped worktrees do NOT persist).
- * - Returns `{}` when `worktreeOpt` is undefined/false (preserves
- *   pre-λ behaviour at non-worktree call sites).
- * - Errors propagate as `WorktreeError`/`Error`; the caller wraps them
- *   in a `createErrorResponse` envelope. Do NOT swallow.
- *
- * Spec: docs/plans/slice-lambda.spec.md §"Implementation surface to
- * verify" §5.
- */
 export declare function resolveWorktreeForRequest(worktreeOpt: boolean | {
     name?: string;
     ref?: string;
 } | undefined, sessionId: string | undefined, runtime: GatewayServerRuntime): Promise<ResolvedWorktree>;
-/**
- * Slice λ §1.d: response-envelope shape decision for `worktreePath`.
- *
- * We surface the worktree path inline as a stdout prefix
- * (`[gateway] worktree=<absolute-path>\n`) rather than as a
- * structuredContent field or JSON wrapper. Rationale:
- *   - zero schema change across all 10 tools and their downstream parsers
- *   - matches how other slice features (session warnings, cache_state
- *     aggregates) surface side-channel metadata today
- *   - callers that want the path can split on the first newline; callers
- *     that don't care see a single ignorable header line
- *
- * Use `formatWorktreePrefix(resolution.worktreePath)` once per tool, at
- * the moment a successful response is constructed.
- */
 export declare function formatWorktreePrefix(worktreePath?: string): string;
-export declare function extractUsageAndCost(cli: "claude" | "codex" | "gemini" | "grok" | "mistral", output: string, outputFormat?: string, 
-/**
- * Optional context for off-stdout telemetry sources. Today only Mistral
- * uses this — its meta.json lives on disk keyed by sessionId. Threading
- * this in keeps the closure built by `buildAsyncFlightRecorderHandoff`
- * primitives-only (no `params`/`prep` retention on AsyncJobRecord).
- */
-ctx?: {
+export declare function extractUsageAndCost(cli: "claude" | "codex" | "gemini" | "grok" | "mistral", output: string, outputFormat?: string, ctx?: {
     sessionId?: string;
     home?: string;
 }): {
@@ -205,35 +115,10 @@ interface CliRequestPrep {
     approvalDecision: ApprovalRecord | null;
     reviewIntegrity?: ReviewIntegrityResult;
     args: string[];
-    /**
-     * Sha256 of the assembled prompt's stable prefix bytes when the caller
-     * supplied `promptParts`. Null when the legacy `prompt` field was used.
-     * Populated by `resolvePromptOrPartsForPrep` and threaded into the
-     * flight-recorder row by the caller's safeFlightStart entry.
-     */
     stablePrefixHash: string | null;
-    /** Heuristic token count (bytes/4) of the same stable prefix. */
     stablePrefixTokens: number | null;
-    /**
-     * Slice κ (Claude only): JSON stream-json payload to feed on stdin
-     * when the gateway emits `-p --input-format stream-json`. Undefined
-     * when the caller did not opt into Anthropic `cache_control`
-     * breakpoints. Non-κ providers always leave this undefined.
-     */
     stdinPayload?: string;
-    /**
-     * Slice κ (Claude only): number of caller-supplied content blocks
-     * that carry an explicit `cache_control` marker. Threaded into the
-     * flight recorder so `cache_state` aggregates can distinguish
-     * κ-explicit breakpoints from implicit prefix-cache hits.
-     */
     cacheControlBlocks?: number;
-    /**
-     * Rec #4: structured warnings produced during prep (e.g. cacheable
-     * stable prefix without cacheControl). Handlers merge these with any
-     * other warnings (cache_ttl_expiring_soon, etc.) before returning to
-     * the caller.
-     */
     warnings?: WarningEntry[];
 }
 export declare function prepareClaudeRequest(params: {
@@ -264,14 +149,12 @@ export declare function prepareClaudeRequest(params: {
     fallbackModel?: string;
     jsonSchema?: string | Record<string, unknown>;
     addDir?: string[];
+    noSessionPersistence?: boolean;
+    settingSources?: string;
+    settings?: string;
+    tools?: string[];
 }, runtime?: GatewayServerRuntime): CliRequestPrep | ExtendedToolResponse;
 export interface CodexRequestPrep extends CliRequestPrep {
-    /**
-     * U26: Cleanup hook for any `outputSchema` temp file written during prep.
-     * Callers MUST invoke this in a `finally` block (regardless of whether the
-     * spawn succeeded, failed, or never ran) to avoid leaking the 0o600 temp
-     * file into `os.tmpdir()`.
-     */
     cleanup?: () => void;
 }
 export declare function prepareCodexRequest(params: {
@@ -292,11 +175,6 @@ export declare function prepareCodexRequest(params: {
     correlationId?: string;
     optimizePrompt: boolean;
     operation: string;
-    /**
-     * U23: output format. When set to "json", emits `--json` so Codex streams
-     * the JSONL event format that `parseCodexJsonStream` (and downstream
-     * `extractUsageAndCost`) can consume. Defaults to "text".
-     */
     outputFormat?: "text" | "json";
     outputSchema?: string | Record<string, unknown>;
     search?: boolean;
@@ -322,24 +200,13 @@ export declare function prepareGeminiRequest(params: {
     correlationId?: string;
     optimizePrompt: boolean;
     operation: string;
-    /**
-     * U23 + Phase 4 slice ε: output format. `json` emits `-o json` (single
-     * JSON object with usageMetadata). `stream-json` emits `-o stream-json`
-     * (NDJSON event stream — `init` / `message` / `result` lines). Both
-     * route through `extractUsageAndCost` so usage tokens reach the flight
-     * recorder. Defaults to "text".
-     */
     outputFormat?: "text" | "json" | "stream-json";
     sandbox?: boolean;
     policyFiles?: string[];
     adminPolicyFiles?: string[];
     attachments?: string[];
-    /**
-     * Phase 4 slice γ: emit `--skip-trust` so first-run workspaces don't
-     * block headless invocations on the interactive trust prompt. Default
-     * is undefined (preserves current prompt behaviour for legacy callers).
-     */
     skipTrust?: boolean;
+    yolo?: boolean;
 }, runtime?: GatewayServerRuntime): CliRequestPrep | ExtendedToolResponse;
 export declare function prepareGrokRequest(params: {
     prompt?: string;
@@ -358,36 +225,15 @@ export declare function prepareGrokRequest(params: {
     correlationId?: string;
     optimizePrompt: boolean;
     operation: string;
-    /**
-     * Phase 4 slice δ: emit `--max-turns N` so callers can cap agent-loop
-     * iterations for cost / latency control. Mirrors Claude's wiring.
-     */
     maxTurns?: number;
-    /**
-     * Phase 4 slice ζ: emit `--cwd <DIR>` so headless callers can set Grok's
-     * working directory without depending on the gateway process's cwd.
-     */
     workingDir?: string;
-    /**
-     * Phase 4 slice θ — Grok HIGH parity. All five are passthrough flags:
-     *
-     * - `sandbox` → `--sandbox <PROFILE>` (freeform; Grok 0.1.210 --help
-     *   shows no enum constraint, unlike --effort / --permission-mode /
-     *   --output-format which all show `[possible values: …]`).
-     * - `rules` → `--rules <RULES>`. Supports `@file` prefix; gateway
-     *   passes the value verbatim and lets Grok parse it.
-     * - `systemPromptOverride` → `--system-prompt-override <PROMPT>`.
-     *   Distinct from Claude's --system-prompt / --append-system-prompt
-     *   (Grok has only one override flag).
-     * - `allow` / `deny` → repeatable `--allow <RULE>` / `--deny <RULE>`
-     *   per --help ("Repeat to add multiple rules"). One argv pair per
-     *   entry — NOT comma-joined like --tools / --disallowed-tools.
-     */
     sandbox?: string;
     rules?: string;
     systemPromptOverride?: string;
     allow?: string[];
     deny?: string[];
+    compactionMode?: string;
+    compactionDetail?: string;
 }, runtime?: GatewayServerRuntime): CliRequestPrep | ExtendedToolResponse;
 export declare function prepareMistralRequest(params: {
     prompt?: string;
@@ -395,8 +241,6 @@ export declare function prepareMistralRequest(params: {
     model?: string;
     outputFormat?: string;
     permissionMode?: MistralAgentMode;
-    effort?: string;
-    reasoningEffort?: string;
     allowedTools?: string[];
     disallowedTools?: string[];
     approvalStrategy: "legacy" | "mcp_managed";
@@ -405,34 +249,16 @@ export declare function prepareMistralRequest(params: {
     correlationId?: string;
     optimizePrompt: boolean;
     operation: string;
-    /**
-     * Phase 4 slice γ: emit `--trust` to bypass Vibe's interactive trust
-     * prompt for this invocation only (not persisted). Default undefined.
-     */
     trust?: boolean;
-    /** Phase 4 slice δ: Vibe `--max-turns N` cap on agent-loop iterations. */
     maxTurns?: number;
-    /** Phase 4 slice δ: Vibe `--max-price DOLLARS` cumulative-cost cap. */
     maxPrice?: number;
-    /** Vibe 2.x: `--max-tokens N` cumulative prompt + completion token cap. */
     maxTokens?: number;
-    /** Phase 4 slice ζ: Vibe `--workdir <DIR>` working-directory parity. */
     workingDir?: string;
-    /** Phase 4 slice ζ: Vibe `--add-dir <DIR>` repeatable add-dir parity. */
     addDir?: string[];
 }, runtime?: GatewayServerRuntime): (CliRequestPrep & {
     mistralEnv: Record<string, string>;
 }) | ExtendedToolResponse;
-/**
- * Phase 4 slice δ post-review: pure helper extracted from
- * `handleMistralRequest` so the retry-path arg-preservation invariants
- * (trust + maxTurns + maxPrice from slices γ/δ) are unit-testable
- * without mocking awaitJobOrDefer. Any param the wrapper threads into
- * the FIRST `buildMistralCliInvocation` call MUST also be threaded
- * through here, or a fresh-workspace / budgeted run can degrade on
- * the second attempt.
- */
-export declare function buildMistralRetryPrep(params: Pick<MistralRequestParams, "outputFormat" | "permissionMode" | "effort" | "reasoningEffort" | "allowedTools" | "disallowedTools" | "approvalStrategy" | "trust" | "maxTurns" | "maxPrice" | "maxTokens" | "workingDir" | "addDir"> & {
+export declare function buildMistralRetryPrep(params: Pick<MistralRequestParams, "outputFormat" | "permissionMode" | "allowedTools" | "disallowedTools" | "approvalStrategy" | "trust" | "maxTurns" | "maxPrice" | "maxTokens" | "workingDir" | "addDir"> & {
     effectivePrompt: string;
 }, recoveryModel: string): {
     args: string[];
@@ -457,18 +283,13 @@ export interface GeminiRequestParams {
     optimizeResponse?: boolean;
     idleTimeoutMs?: number;
     forceRefresh?: boolean;
-    /**
-     * U23 + Phase 4 slice ε: "json" emits `-o json`; "stream-json" emits
-     * `-o stream-json` (NDJSON event stream). Both are usage-extracted.
-     */
     outputFormat?: "text" | "json" | "stream-json";
     sandbox?: boolean;
     policyFiles?: string[];
     adminPolicyFiles?: string[];
     attachments?: string[];
-    /** Phase 4 slice γ: emit `--skip-trust` for fresh-workspace headless runs. */
     skipTrust?: boolean;
-    /** Slice λ: run this request inside a gateway-owned git worktree. */
+    yolo?: boolean;
     worktree?: boolean | {
         name?: string;
         ref?: string;
@@ -511,21 +332,15 @@ export interface GrokRequestParams {
     optimizeResponse?: boolean;
     idleTimeoutMs?: number;
     forceRefresh?: boolean;
-    /** Phase 4 slice δ: cap agent-loop iterations via `--max-turns N`. */
     maxTurns?: number;
-    /** Phase 4 slice ζ: emit `--cwd <DIR>` so the CLI uses the specified working directory. */
     workingDir?: string;
-    /** Phase 4 slice θ: Grok `--sandbox <PROFILE>` (freeform passthrough). */
     sandbox?: string;
-    /** Phase 4 slice θ: Grok `--rules <RULES>` (supports `@file` prefix; verbatim passthrough). */
     rules?: string;
-    /** Phase 4 slice θ: Grok `--system-prompt-override <PROMPT>`. */
     systemPromptOverride?: string;
-    /** Phase 4 slice θ: Grok `--allow <RULE>` (repeatable; one entry per --allow instance). */
     allow?: string[];
-    /** Phase 4 slice θ: Grok `--deny <RULE>` (repeatable; one entry per --deny instance). */
     deny?: string[];
-    /** Slice λ: run this request inside a gateway-owned git worktree. */
+    compactionMode?: string;
+    compactionDetail?: string;
     worktree?: boolean | {
         name?: string;
         ref?: string;
@@ -542,8 +357,6 @@ export interface MistralRequestParams {
     resumeLatest: boolean;
     createNewSession: boolean;
     permissionMode?: MistralAgentMode;
-    effort?: string;
-    reasoningEffort?: string;
     approvalStrategy: "legacy" | "mcp_managed";
     approvalPolicy?: string;
     mcpServers?: ClaudeMcpServerName[];
@@ -554,19 +367,12 @@ export interface MistralRequestParams {
     optimizeResponse?: boolean;
     idleTimeoutMs?: number;
     forceRefresh?: boolean;
-    /** Phase 4 slice γ: emit `--trust` for fresh-workspace headless runs. */
     trust?: boolean;
-    /** Phase 4 slice δ: Vibe `--max-turns N` cap on agent-loop iterations. */
     maxTurns?: number;
-    /** Phase 4 slice δ: Vibe `--max-price DOLLARS` cumulative-cost cap. */
     maxPrice?: number;
-    /** Vibe 2.x: `--max-tokens N` cumulative prompt + completion token cap. */
     maxTokens?: number;
-    /** Phase 4 slice ζ: Vibe `--workdir <DIR>` working-directory parity. */
     workingDir?: string;
-    /** Phase 4 slice ζ: Vibe `--add-dir <DIR>` repeatable add-dir parity. */
     addDir?: string[];
-    /** Slice λ: run this request inside a gateway-owned git worktree. */
     worktree?: boolean | {
         name?: string;
         ref?: string;
@@ -593,7 +399,6 @@ export declare function handleCodexRequestAsync(deps: AsyncHandlerDeps, params:
     optimizePrompt: boolean;
     idleTimeoutMs?: number;
     forceRefresh?: boolean;
-    /** U23: when "json", emits Codex `--json` so the parser is reachable. */
     outputFormat?: "text" | "json";
     outputSchema?: string | Record<string, unknown>;
     search?: boolean;
@@ -605,7 +410,6 @@ export declare function handleCodexRequestAsync(deps: AsyncHandlerDeps, params:
     ignoreRules?: boolean;
     workingDir?: string;
     addDir?: string[];
-    /** Slice λ: run this request inside a gateway-owned git worktree. */
     worktree?: boolean | {
         name?: string;
         ref?: string;