llm-cli-gateway 1.15.3 → 1.16.1

package/dist/health.js

@@ -1,9 +1,6 @@
 import { listProviderRuntimeStatuses } from "./provider-status.js";
 /**
- * Check health status of PostgreSQL and Redis
- * - Both up → healthy
- * - Only PostgreSQL up → degraded (Redis down but DB works)
- * - PostgreSQL down → unhealthy (critical failure)
+ * Check health status of PostgreSQL.
  */
 export async function checkHealth(db) {
     const result = await db.healthCheck();
@@ -13,22 +10,9 @@ export async function checkHealth(db) {
             status: result.postgres.connected ? "up" : "down",
             latency: result.postgres.latency,
         },
-        redis: {
-            status: result.redis.connected ? "up" : "down",
-            latency: result.redis.latency,
-        },
         timestamp: new Date().toISOString(),
     };
-    // Determine overall health status
-    if (result.postgres.connected && result.redis.connected) {
-        health.status = "healthy";
-    }
-    else if (result.postgres.connected && !result.redis.connected) {
-        health.status = "degraded";
-    }
-    else {
-        health.status = "unhealthy";
-    }
+    health.status = result.postgres.connected ? "healthy" : "unhealthy";
     return health;
 }
 export function checkProviderRuntimeHealth() {

package/dist/index.d.ts

@@ -413,6 +413,8 @@ export declare function prepareMistralRequest(params: {
     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. */
@@ -429,7 +431,7 @@ export declare function prepareMistralRequest(params: {
  * 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" | "workingDir" | "addDir"> & {
+export declare function buildMistralRetryPrep(params: Pick<MistralRequestParams, "outputFormat" | "permissionMode" | "effort" | "reasoningEffort" | "allowedTools" | "disallowedTools" | "approvalStrategy" | "trust" | "maxTurns" | "maxPrice" | "maxTokens" | "workingDir" | "addDir"> & {
     effectivePrompt: string;
 }, recoveryModel: string): {
     args: string[];
@@ -557,6 +559,8 @@ export interface MistralRequestParams {
     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. */

package/dist/index.js

@@ -165,7 +165,7 @@ Other: list_models, cli_versions, upstream_contracts, cli_upgrade, approval_list
 
 Key behaviors:
 - Sync auto-defers at ${SYNC_DEADLINE_MS}ms. Poll deferred jobs via llm_job_status/llm_job_result.
-- Sessions: Claude --continue, Gemini --resume, Grok --resume/--continue, Mistral --resume/--continue (requires session_logging.enabled=true in ~/.vibe/config.toml), Codex \`exec resume <ID>\` / \`exec resume --last\` (all real CLI continuity). For Codex, sessionId must be a real Codex UUID (from ~/.codex/sessions/); gateway-generated gw-* IDs are rejected.
+- Sessions: Claude --continue, Gemini --resume, Grok --resume/--continue, Mistral --resume/--continue (current Vibe defaults session logging on; doctor flags explicit session_logging.enabled=false), Codex \`exec resume <ID>\` / \`exec resume --last\` (all real CLI continuity). For Codex, sessionId must be a real Codex UUID (from ~/.codex/sessions/); gateway-generated gw-* IDs are rejected.
 - Approval gates: opt-in via approvalStrategy:"mcp_managed".
 - Idle timeout kills stuck processes (default 10min, configurable via idleTimeoutMs).
 
@@ -1766,6 +1766,7 @@ export function prepareMistralRequest(params, runtime = resolveGatewayServerRunt
         trust: params.trust,
         maxTurns: params.maxTurns,
         maxPrice: params.maxPrice,
+        maxTokens: params.maxTokens,
         workingDir: params.workingDir,
         addDir: params.addDir,
     });
@@ -1822,6 +1823,7 @@ export function buildMistralRetryPrep(params, recoveryModel) {
         trust: params.trust,
         maxTurns: params.maxTurns,
         maxPrice: params.maxPrice,
+        maxTokens: params.maxTokens,
         workingDir: params.workingDir,
         addDir: params.addDir,
     });
@@ -2428,6 +2430,7 @@ export async function handleMistralRequest(deps, params) {
         trust: params.trust,
         maxTurns: params.maxTurns,
         maxPrice: params.maxPrice,
+        maxTokens: params.maxTokens,
         workingDir: params.workingDir,
         addDir: params.addDir,
     }, runtime);
@@ -2578,6 +2581,7 @@ export async function handleMistralRequestAsync(deps, params) {
         trust: params.trust,
         maxTurns: params.maxTurns,
         maxPrice: params.maxPrice,
+        maxTokens: params.maxTokens,
         workingDir: params.workingDir,
         addDir: params.addDir,
     }, runtime);
@@ -3761,13 +3765,13 @@ export function createGatewayServer(deps = {}) {
             .optional()
             .describe("Model alias (e.g. mistral-medium-3.5, latest). Resolved alias is injected via VIBE_ACTIVE_MODEL env var; Vibe has no --model flag."),
         outputFormat: z
-            .enum(["plain", "json", "stream-json"])
+            .enum(["text", "plain", "json", "streaming", "stream-json"])
             .optional()
-            .describe("Output format (plain|json|stream-json). Vibe default is plain."),
+            .describe("Output format for Vibe 2.x (text|json|streaming). Legacy aliases plain→text and stream-json→streaming are accepted."),
         sessionId: z
             .string()
             .optional()
-            .describe("Session ID (user-provided CLI handle for --resume). Requires [session_logging] enabled = true in ~/.vibe/config.toml."),
+            .describe("Session ID (user-provided CLI handle for --resume). Current Vibe defaults session logging on; doctor flags explicit [session_logging] enabled = false."),
         resumeLatest: z
             .boolean()
             .default(false)
@@ -3822,6 +3826,7 @@ export function createGatewayServer(deps = {}) {
             .describe("Emit `--trust` so Vibe trusts the cwd for this invocation only (not persisted to trusted_folders.toml) and skips the interactive trust prompt (Phase 4 slice γ)."),
         maxTurns: MAX_TURNS_SCHEMA.optional().describe("Vibe `--max-turns N`: cap the agent-loop iteration count (programmatic mode only, Phase 4 slice δ). Bounded to safe integers ≤ 10000."),
         maxPrice: MAX_PRICE_SCHEMA.optional().describe("Vibe `--max-price DOLLARS`: interrupt the session when cumulative cost crosses this cap (programmatic mode only, Phase 4 slice δ). Bounded to finite values ≤ 10000 USD."),
+        maxTokens: MAX_TURNS_SCHEMA.optional().describe("Vibe `--max-tokens N`: cap cumulative prompt + completion tokens for the session (programmatic mode only). Bounded to safe integers ≤ 10000."),
         // Phase 4 slice ζ — Vibe working-directory + additional-dirs parity.
         workingDir: z
             .string()
@@ -3833,7 +3838,7 @@ export function createGatewayServer(deps = {}) {
             .optional()
             .describe("Vibe --add-dir <DIR>: additional writable workspace directories. Each entry is emitted as its own --add-dir instance (Vibe states this flag may be specified multiple times)."),
         worktree: WORKTREE_SCHEMA.optional(),
-    }, async ({ prompt, promptParts, model, outputFormat, sessionId, resumeLatest, createNewSession, permissionMode, effort, reasoningEffort, approvalStrategy, approvalPolicy, mcpServers, allowedTools, disallowedTools, correlationId, optimizePrompt, optimizeResponse, idleTimeoutMs, forceRefresh, trust, maxTurns, maxPrice, workingDir, addDir, worktree, }) => {
+    }, async ({ prompt, promptParts, model, outputFormat, sessionId, resumeLatest, createNewSession, permissionMode, effort, reasoningEffort, approvalStrategy, approvalPolicy, mcpServers, allowedTools, disallowedTools, correlationId, optimizePrompt, optimizeResponse, idleTimeoutMs, forceRefresh, trust, maxTurns, maxPrice, maxTokens, workingDir, addDir, worktree, }) => {
         return handleMistralRequest({ sessionManager, logger, runtime }, {
             prompt,
             promptParts,
@@ -3858,6 +3863,7 @@ export function createGatewayServer(deps = {}) {
             trust,
             maxTurns,
             maxPrice,
+            maxTokens,
             workingDir,
             addDir,
             worktree,
@@ -4482,13 +4488,13 @@ export function createGatewayServer(deps = {}) {
                 .optional()
                 .describe("Model alias (resolved into VIBE_ACTIVE_MODEL env var — Vibe has no --model flag)"),
             outputFormat: z
-                .enum(["plain", "json", "stream-json"])
+                .enum(["text", "plain", "json", "streaming", "stream-json"])
                 .optional()
-                .describe("Output format (plain|json|stream-json). Vibe default is plain."),
+                .describe("Output format for Vibe 2.x (text|json|streaming). Legacy aliases plain→text and stream-json→streaming are accepted."),
             sessionId: z
                 .string()
                 .optional()
-                .describe("Session ID (user-provided CLI handle for --resume). Requires [session_logging] enabled = true in ~/.vibe/config.toml."),
+                .describe("Session ID (user-provided CLI handle for --resume). Current Vibe defaults session logging on; doctor flags explicit [session_logging] enabled = false."),
             resumeLatest: z
                 .boolean()
                 .default(false)
@@ -4542,6 +4548,7 @@ export function createGatewayServer(deps = {}) {
                 .describe("Emit `--trust` so Vibe trusts the cwd for this invocation only (not persisted to trusted_folders.toml) and skips the interactive trust prompt (Phase 4 slice γ)."),
             maxTurns: MAX_TURNS_SCHEMA.optional().describe("Vibe `--max-turns N`: cap the agent-loop iteration count (programmatic mode only, Phase 4 slice δ). Bounded to safe integers ≤ 10000."),
             maxPrice: MAX_PRICE_SCHEMA.optional().describe("Vibe `--max-price DOLLARS`: interrupt the session when cumulative cost crosses this cap (programmatic mode only, Phase 4 slice δ). Bounded to finite values ≤ 10000 USD."),
+            maxTokens: MAX_TURNS_SCHEMA.optional().describe("Vibe `--max-tokens N`: cap cumulative prompt + completion tokens for the session (programmatic mode only). Bounded to safe integers ≤ 10000."),
             // Phase 4 slice ζ — Vibe working-directory + additional-dirs parity.
             workingDir: z
                 .string()
@@ -4553,7 +4560,7 @@ export function createGatewayServer(deps = {}) {
                 .optional()
                 .describe("Vibe --add-dir <DIR>: additional writable workspace directories. Each entry is emitted as its own --add-dir instance."),
             worktree: WORKTREE_SCHEMA.optional(),
-        }, async ({ prompt, promptParts, model, outputFormat, sessionId, resumeLatest, createNewSession, permissionMode, effort, reasoningEffort, approvalStrategy, approvalPolicy, mcpServers, allowedTools, disallowedTools, correlationId, optimizePrompt, idleTimeoutMs, forceRefresh, trust, maxTurns, maxPrice, workingDir, addDir, worktree, }) => {
+        }, async ({ prompt, promptParts, model, outputFormat, sessionId, resumeLatest, createNewSession, permissionMode, effort, reasoningEffort, approvalStrategy, approvalPolicy, mcpServers, allowedTools, disallowedTools, correlationId, optimizePrompt, idleTimeoutMs, forceRefresh, trust, maxTurns, maxPrice, maxTokens, workingDir, addDir, worktree, }) => {
             return handleMistralRequestAsync({ sessionManager, asyncJobManager, logger, runtime }, {
                 prompt,
                 promptParts,
@@ -4577,6 +4584,7 @@ export function createGatewayServer(deps = {}) {
                 trust,
                 maxTurns,
                 maxPrice,
+                maxTokens,
                 workingDir,
                 addDir,
                 worktree,
@@ -5101,8 +5109,8 @@ async function initializeSessionManager() {
     // own a single filesystem); revisit if/when worktree support extends
     // there.
     const worktreeCleanupHook = createWorktreeSessionCleanupHook(logger);
-    if (config.database && config.redis) {
-        logger.info("Initializing PostgreSQL + Redis session manager");
+    if (config.database) {
+        logger.info("Initializing PostgreSQL session manager");
         const { createDatabaseConnection } = await import("./db.js");
         db = await createDatabaseConnection(config, logger);
         sessionManager = await createSessionManager(config, db, logger);

package/dist/migrate-sessions.js

@@ -82,7 +82,6 @@ Options:
 
 Environment Variables:
   DATABASE_URL     PostgreSQL connection string (required)
-  REDIS_URL        Redis connection string (required)
 `);
         process.exit(args[0] === "--help" || args[0] === "-h" ? 0 : 1);
     }
@@ -97,18 +96,17 @@ Environment Variables:
     console.error(`Migration Configuration:`);
     console.error(`  Source: ${filePath}`);
     console.error(`  DATABASE_URL: ${process.env.DATABASE_URL ? "[set]" : "[not set]"}`);
-    console.error(`  REDIS_URL: ${process.env.REDIS_URL ? "[set]" : "[not set]"}`);
     console.error("");
     // Load config
     const config = loadConfig();
-    if (!config.database || !config.redis) {
-        console.error("ERROR: DATABASE_URL and REDIS_URL must be set");
+    if (!config.database) {
+        console.error("ERROR: DATABASE_URL must be set");
         process.exit(1);
     }
     // Connect to database
     console.error("Connecting to database...");
     const db = await createDatabaseConnection(config, logger);
-    const pgManager = new PostgreSQLSessionManager(db.getPool(), db.getRedis(), config.cacheTtl, logger);
+    const pgManager = new PostgreSQLSessionManager(db.getPool());
     console.error("✓ Connected to database\n");
     try {
         // Run migration

package/dist/provider-login-guidance.js

@@ -76,17 +76,22 @@ const GUIDANCE = {
         displayName: "Mistral Vibe CLI",
         install: {
             summary: "Install Mistral Vibe CLI via pip, uv, or Homebrew (Vibe does not self-update; cli_upgrade dispatches to the installer it detects).",
-            commands: ["pip install vibe-cli", "uv tool install vibe-cli", "brew install mistral-vibe"],
-            documentationUrl: "https://docs.mistral.ai/agents/vibe-cli",
+            commands: [
+                "curl -LsSf https://mistral.ai/vibe/install.sh | bash",
+                "pip install mistral-vibe",
+                "uv tool install mistral-vibe",
+                "brew install mistral-vibe",
+            ],
+            documentationUrl: "https://docs.mistral.ai/mistral-vibe/overview",
         },
         login: {
-            summary: "Sign in through Mistral's official auth flow and enable session_logging in ~/.vibe/config.toml.",
-            commands: ["vibe auth login", "vibe config set session_logging.enabled true"],
+            summary: "Sign in through Mistral's official auth flow. Current Vibe defaults session logging to enabled; if an older config disabled it, edit ~/.vibe/config.toml and set [session_logging] enabled = true.",
+            commands: ["vibe auth login"],
             credentialHandling: "Do not paste Mistral API keys, OAuth tokens, or ~/.vibe/credentials into the gateway or a remote chat.",
         },
         verification: {
             command: "vibe --version",
-            expected: "Vibe CLI is installed; doctor additionally checks ~/.vibe/config.toml for session_logging.enabled=true",
+            expected: "Vibe CLI is installed; doctor checks ~/.vibe/config.toml for an explicit session_logging.enabled=false override",
         },
     },
 };

package/dist/provider-status.js

@@ -12,7 +12,7 @@ const VERSION_ARGS = {
     grok: ["--version"],
     mistral: ["--version"],
 };
-// Mistral Vibe ships as the `vibe` binary (PyPI package vibe-cli); the gateway
+// Mistral Vibe ships as the `vibe` binary (PyPI package mistral-vibe); the gateway
 // uses `mistral` as the provider key but invokes `vibe` on the shell.
 export const PROVIDER_COMMANDS = {
     claude: "claude",

package/dist/request-helpers.d.ts

@@ -71,9 +71,9 @@ export declare function resolveGrokSessionArgs(opts: {
 /**
  * Mistral Vibe-specific resume args.
  *
- * Vibe persists sessions only when `[session_logging] enabled = true` is set in
- * `~/.vibe/config.toml`. The doctor checks for that toggle and surfaces an
- * actionable error when it is missing; this pure helper just emits the args.
+ * Current Vibe defaults session logging on; older configs can explicitly set
+ * `[session_logging] enabled = false`. The doctor checks that toggle before
+ * callers rely on session continuity; this pure helper just emits the args.
  *
  * The args shape mirrors Grok (`--continue` for latest, `--resume <id>` for a
  * specific session) because Vibe exposes the same surface for its session log.
@@ -125,6 +125,11 @@ export interface PrepareMistralRequestInput {
      * only).
      */
     maxPrice?: number;
+    /**
+     * Vibe 2.x supports `--max-tokens N` in programmatic mode, wired through to
+     * `run_programmatic(max_session_tokens=...)`.
+     */
+    maxTokens?: number;
     /**
      * Phase 4 slice ζ: emit `--workdir <DIR>` so Vibe changes into the named
      * directory before running. Single value (Vibe accepts one --workdir).
@@ -148,6 +153,8 @@ export interface PrepareMistralRequestResult {
  * - Model is selected via `VIBE_ACTIVE_MODEL` env var (NOT a `--model` flag).
  * - Permission mode emits `--agent <mode>` (defaults to `auto-approve` when unset).
  * - Allowed tools emit `--enabled-tools <tool>` once per tool (allowlist only).
+ * - Output format emits `--output <text|json|streaming>` (legacy gateway
+ *   aliases `plain` and `stream-json` are normalized before spawn).
  * - Disallowed tools are accepted but ignored at the CLI boundary.
  */
 export declare function prepareMistralRequest(input: PrepareMistralRequestInput): PrepareMistralRequestResult;

package/dist/request-helpers.js

@@ -100,9 +100,9 @@ export function resolveGrokSessionArgs(opts) {
 /**
  * Mistral Vibe-specific resume args.
  *
- * Vibe persists sessions only when `[session_logging] enabled = true` is set in
- * `~/.vibe/config.toml`. The doctor checks for that toggle and surfaces an
- * actionable error when it is missing; this pure helper just emits the args.
+ * Current Vibe defaults session logging on; older configs can explicitly set
+ * `[session_logging] enabled = false`. The doctor checks that toggle before
+ * callers rely on session continuity; this pure helper just emits the args.
  *
  * The args shape mirrors Grok (`--continue` for latest, `--resume <id>` for a
  * specific session) because Vibe exposes the same surface for its session log.
@@ -151,6 +151,8 @@ export const MISTRAL_DEFAULT_AGENT_MODE = "auto-approve";
  * - Model is selected via `VIBE_ACTIVE_MODEL` env var (NOT a `--model` flag).
  * - Permission mode emits `--agent <mode>` (defaults to `auto-approve` when unset).
  * - Allowed tools emit `--enabled-tools <tool>` once per tool (allowlist only).
+ * - Output format emits `--output <text|json|streaming>` (legacy gateway
+ *   aliases `plain` and `stream-json` are normalized before spawn).
  * - Disallowed tools are accepted but ignored at the CLI boundary.
  */
 export function prepareMistralRequest(input) {
@@ -160,7 +162,7 @@ export function prepareMistralRequest(input) {
         env.VIBE_ACTIVE_MODEL = input.resolvedModel;
     }
     if (input.outputFormat) {
-        args.push("--output-format", input.outputFormat);
+        args.push("--output", normalizeMistralOutputFormat(input.outputFormat));
     }
     const mode = input.permissionMode ?? MISTRAL_DEFAULT_AGENT_MODE;
     args.push("--agent", mode);
@@ -185,6 +187,9 @@ export function prepareMistralRequest(input) {
     if (input.maxPrice !== undefined) {
         args.push("--max-price", String(input.maxPrice));
     }
+    if (input.maxTokens !== undefined) {
+        args.push("--max-tokens", String(input.maxTokens));
+    }
     if (input.workingDir) {
         args.push("--workdir", input.workingDir);
     }
@@ -196,6 +201,13 @@ export function prepareMistralRequest(input) {
     const ignoredDisallowedTools = Boolean(input.disallowedTools && input.disallowedTools.length > 0);
     return { args, env, ignoredDisallowedTools };
 }
+function normalizeMistralOutputFormat(format) {
+    if (format === "plain")
+        return "text";
+    if (format === "stream-json")
+        return "streaming";
+    return format;
+}
 //──────────────────────────────────────────────────────────────────────────────
 // U24: Permission / approval mode parity helpers
 //──────────────────────────────────────────────────────────────────────────────

package/dist/session-manager-pg.d.ts

@@ -1,77 +1,48 @@
 import type { Pool } from "pg";
-import type { Redis } from "ioredis";
 import { Session, CliType } from "./session-manager.js";
-import { CacheTtl } from "./config.js";
-import type { Logger } from "./logger.js";
 export type { Logger } from "./logger.js";
 /**
- * PostgreSQL-backed session manager with Redis caching
+ * PostgreSQL-backed session manager. PostgreSQL is the source of truth and
+ * the only required service for this backend.
  */
 export declare class PostgreSQLSessionManager {
     private pool;
-    private redis;
-    private cacheTtl;
-    private logger;
-    constructor(pool: Pool, redis: Redis, cacheTtl: CacheTtl, logger: Logger);
+    constructor(pool: Pool);
     /**
-     * Acquire distributed lock using Redis SET NX EX
-     * Returns [success, lockValue] tuple
-     */
-    private acquireLock;
-    private sleep;
-    /**
-     * Acquire a distributed lock with bounded retries to smooth contention spikes.
-     */
-    private acquireLockWithRetry;
-    /**
-     * Release distributed lock with optimistic Redis transaction semantics.
-     * Only releases if lockValue matches, which prevents releasing another
-     * process's lock after expiry/reacquire.
-     */
-    private releaseLock;
-    /**
-     * Invalidate session cache
-     */
-    private invalidateCache;
-    /**
-     * Invalidate session list cache using SCAN (non-blocking)
-     */
-    private invalidateListCache;
-    /**
-     * Create a new session
+     * Create a new session.
      */
     createSession(cli: CliType, description?: string, sessionId?: string): Promise<Session>;
     /**
-     * Get session by ID (cache-aside pattern)
+     * Get session by ID.
      */
     getSession(sessionId: string): Promise<Session | null>;
     /**
-     * List all sessions, optionally filtered by CLI
+     * List all sessions, optionally filtered by CLI.
      */
     listSessions(cli?: CliType): Promise<Session[]>;
     /**
-     * Delete a session
+     * Delete a session.
      */
     deleteSession(sessionId: string): Promise<boolean>;
     /**
-     * Set active session for a CLI (with distributed locking)
+     * Set active session for a CLI. The row-level update is serialized by
+     * PostgreSQL and the session FK keeps stale IDs from being recorded.
      */
     setActiveSession(cli: CliType, sessionId: string | null): Promise<boolean>;
     /**
-     * Get active session for a CLI
+     * Get active session for a CLI.
      */
     getActiveSession(cli: CliType): Promise<Session | null>;
     /**
-     * Update session usage timestamp
+     * Update session usage timestamp.
      */
     updateSessionUsage(sessionId: string): Promise<void>;
     /**
-     * Update session metadata (atomic JSONB merge)
+     * Update session metadata using PostgreSQL's atomic JSONB merge.
      */
     updateSessionMetadata(sessionId: string, metadata: Record<string, any>): Promise<boolean>;
     /**
-     * Clear all sessions, optionally filtered by CLI
-     * Invalidates all related caches (session, active, list)
+     * Clear all sessions, optionally filtered by CLI.
      */
     clearAllSessions(cli?: CliType): Promise<number>;
 }