@2kw/ai 6.2.0-dev.25 → 6.2.0-dev.30

package/dist/commands/agent-run.js

@@ -52,10 +52,10 @@ async function withAgentSuggestions(client, agentRef, err) {
 /**
  * Prints a run result and sets the exit code. In interactive mode an approval pause
  * is decided at the terminal and continued until the run stops pausing for approval.
- * Continuations stay on the run's label, and the envelope's `next` names it.
+ * Continuations stay on the run's label and model override, and the envelope's `next` names both.
  */
 async function finishRun(command, client, target, first, opts) {
-    const agentRef = target.label ? `${target.name}@${target.label}` : target.name;
+    const agentRef = `${target.name}${target.label ? `@${target.label}` : ""}${target.model ? `#${target.model}` : ""}`;
     let result = first;
     let env = buildRunEnvelope(result, agentRef);
     if (opts.interactive && env.status === "requires_approval") {
@@ -90,7 +90,7 @@ async function finishRun(command, client, target, first, opts) {
                     throw new CliUsageError(`No pending approvals for ${responseId} (already decided or superseded).`);
                 }
                 const decisions = decisionsFromAnswers(rows, answers);
-                result = await withSpinner("Continuing...", () => continueWithDecisions(client, String(agent.id), responseId, decisions, target.label));
+                result = await withSpinner("Continuing...", () => continueWithDecisions(client, String(agent.id), responseId, decisions, target.label, target.model));
                 env = buildRunEnvelope(result, agentRef);
             }
         }
@@ -116,7 +116,7 @@ async function finishRun(command, client, target, first, opts) {
 export function makeAgentRunCommand() {
     return new Command("run")
         .description("Run an agent on a task (exit 3 = paused for approval, 4 = needs client tool output, 5 = incomplete)")
-        .argument("<agent>", "Agent name or id, optionally with @label")
+        .argument("<agent>", "Agent name or id, optionally with @label and #model")
         .argument("[input]", "Task text (or use --input-file, or pipe it on stdin)")
         .option("--input-file <path>", "Read the task text from a file")
         .option("--conversation <id>", "Continue a conversation")
@@ -145,8 +145,8 @@ export function makeAgentRunCommand() {
             }
             // The prompt reads stdin and writes to stderr, so those two decide; stdout may be redirected.
             const interactive = !!process.stdin.isTTY && !!process.stderr.isTTY && !isJsonOutput(command) && opts.input !== false && !opts.raw;
-            const { name, label } = splitAgentRef(agentRef);
-            await finishRun(command, client, { name, label }, data, {
+            const { name, label, model } = splitAgentRef(agentRef);
+            await finishRun(command, client, { name, label, model }, data, {
                 raw: !!opts.raw,
                 interactive,
             });
@@ -156,7 +156,7 @@ export function makeAgentRunCommand() {
 export function makeAgentDecideCommand() {
     return new Command("decide")
         .description("Decide every pending approval of a paused run and continue it")
-        .argument("<agent>", "Agent name or id, optionally with @label (use the same label the run used)")
+        .argument("<agent>", "Agent name or id, optionally with @label and #model (use the same @label and #model the run used)")
         .requiredOption("--response <responseId>", "The paused response id (responseId in the run envelope)")
         .option("--approve <ids...>", "Approval ids to approve")
         .option("--reject <ids...>", "Approval ids to reject")
@@ -168,7 +168,10 @@ export function makeAgentDecideCommand() {
         .action(async (agentRef, opts, command) => {
         await runAction(command, async () => {
             const client = getClient(command);
-            const { name, label } = splitAgentRef(agentRef);
+            const { name, label, model } = splitAgentRef(agentRef);
+            // Dropped silently, an empty override would continue the run on the default model.
+            if (model === "")
+                throw new CliUsageError(`Empty #model in '${agentRef}': use the model the run used, or drop the '#'.`);
             const agent = await resolveAgent(client, name);
             const pending = await fetchPendingApprovals(client, String(agent.id), opts.response);
             const decisions = planDecisions(pending, {
@@ -179,8 +182,8 @@ export function makeAgentDecideCommand() {
                 reason: opts.reason,
                 remember: opts.remember,
             }, opts.response);
-            const result = await withSpinner("Continuing...", () => continueWithDecisions(client, String(agent.id), opts.response, decisions, label));
-            await finishRun(command, client, { name, label }, result, { raw: !!opts.raw, interactive: false });
+            const result = await withSpinner("Continuing...", () => continueWithDecisions(client, String(agent.id), opts.response, decisions, label, model));
+            await finishRun(command, client, { name, label, model }, result, { raw: !!opts.raw, interactive: false });
         });
     });
 }

package/dist/commands/providers.js

@@ -35,7 +35,7 @@ export function makeProvidersCommand() {
         .command("create")
         .description("Create a new provider")
         .requiredOption("-n, --name <name>", "Provider name")
-        .requiredOption("--type <type>", "Provider type (openai, azure_openai, anthropic, xai, mistral, vertex_ai, ollama)")
+        .requiredOption("--type <type>", "Provider type (openai, azure_openai, anthropic, xai, mistral, ollama)")
         .requiredOption("--provider-api-key <key>", "API key for the AI provider")
         .option("-c, --config <json>", "Provider config as JSON")
         .action(async (opts, command) => {
@@ -107,7 +107,7 @@ export function makeProvidersCommand() {
     cmd
         .command("test")
         .description("Test connection to an AI provider using given or stored credentials")
-        .requiredOption("--type <type>", "Provider type (openai, azure_openai, anthropic, xai, mistral, vertex_ai, ollama)")
+        .requiredOption("--type <type>", "Provider type (openai, azure_openai, anthropic, xai, mistral, ollama)")
         .option("--provider-api-key <key>", "API key to test (omit to use --provider-id's stored key)")
         .option("--provider-id <id>", "Existing provider ID whose stored credentials should be used")
         .option("-c, --config <json>", "Provider config as JSON")

package/dist/lib/agent-decide.d.ts

@@ -32,9 +32,10 @@ export declare function decisionsFromAnswers(pending: ApprovalRow[], answers: Ma
 /** Same item shape as n8n Decide Approval (#660). */
 export declare function toApprovalItems(decisions: Decision[]): Record<string, string>[];
 /**
- * Continues on `agent/<agentId>`, or `agent/<agentId>@<label>` when the run used a label: the server
- * resolves the approved call's tools from the continuation's version, so the label must match the run's.
- * Never the echoed `name@<versionNumber>` (it would 404 as a label) and never a `#model` suffix.
+ * Continues on `agent/<agentId>[@<label>][#<model>]` with the label and model override the run used: the
+ * server resolves the continuation's tools, instructions and model from this reference, so both must match
+ * the run's or the continuation drops to another version or the agent's default model.
+ * Never the echoed `name@<versionNumber>` (it would 404 as a label).
  */
-export declare function continueWithDecisions(client: ApiClient, agentId: string, responseId: string, decisions: Decision[], label?: string): Promise<ResponsesResult>;
+export declare function continueWithDecisions(client: ApiClient, agentId: string, responseId: string, decisions: Decision[], label?: string, model?: string): Promise<ResponsesResult>;
 //# sourceMappingURL=agent-decide.d.ts.map

package/dist/lib/agent-decide.js

@@ -109,13 +109,14 @@ export function toApprovalItems(decisions) {
     }));
 }
 /**
- * Continues on `agent/<agentId>`, or `agent/<agentId>@<label>` when the run used a label: the server
- * resolves the approved call's tools from the continuation's version, so the label must match the run's.
- * Never the echoed `name@<versionNumber>` (it would 404 as a label) and never a `#model` suffix.
+ * Continues on `agent/<agentId>[@<label>][#<model>]` with the label and model override the run used: the
+ * server resolves the continuation's tools, instructions and model from this reference, so both must match
+ * the run's or the continuation drops to another version or the agent's default model.
+ * Never the echoed `name@<versionNumber>` (it would 404 as a label).
  */
-export async function continueWithDecisions(client, agentId, responseId, decisions, label) {
+export async function continueWithDecisions(client, agentId, responseId, decisions, label, model) {
     const body = {
-        model: label ? `agent/${agentId}@${label}` : `agent/${agentId}`,
+        model: `agent/${agentId}${label ? `@${label}` : ""}${model ? `#${model}` : ""}`,
         previous_response_id: responseId,
         stream: false,
         input: toApprovalItems(decisions),

package/dist/lib/agent-lookup.d.ts

@@ -2,6 +2,7 @@ import type { components } from "../generated/openapi.js";
 import type { getClient } from "./client.js";
 export type ApiClient = ReturnType<typeof getClient>;
 export type AgentDTO = components["schemas"]["AgentDTO"];
+/** `name[@label][#model]`, split like the server: the model at the last `#`, then the label at the last `@`. */
 export declare function splitAgentRef(ref: string): {
     name: string;
     label: string | undefined;

package/dist/lib/agent-lookup.js

@@ -3,8 +3,9 @@ import { paginationParams } from "./pagination.js";
 const UUID = /^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/i;
 const PAGE_SIZE = 100;
 const MAX_PAGES = 50;
+/** `name[@label][#model]`, split like the server: the model at the last `#`, then the label at the last `@`. */
 export function splitAgentRef(ref) {
-    const hash = ref.indexOf("#");
+    const hash = ref.lastIndexOf("#");
     const model = hash >= 0 ? ref.slice(hash + 1) : undefined;
     const base = hash >= 0 ? ref.slice(0, hash) : ref;
     const at = base.lastIndexOf("@");

package/dist/lib/agent-run.js

@@ -24,8 +24,8 @@ export function extractResponseText(result) {
  * any other `@` suffix stays part of the name.
  */
 export function parseAgentModel(model) {
-    // Strip the `#<model>` override before matching the agent reference.
-    const hash = (model ?? "").indexOf("#");
+    // Strip the `#<model>` override (from the last `#`, as the server splits it) before matching.
+    const hash = (model ?? "").lastIndexOf("#");
     const ref = hash >= 0 ? (model ?? "").slice(0, hash) : model ?? "";
     const withVersion = /^agent\/(.+)@(\d+)$/.exec(ref);
     if (withVersion)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@2kw/ai",
-  "version": "6.2.0-dev.25",
+  "version": "6.2.0-dev.30",
   "description": "CLI for 2kw.ai — schema-driven document extraction, an OpenAI-compatible EU LLM gateway, transcription, prompts, datasets, and experiments from your terminal or agentic workflows. Ships as 2kw, backbone, and bb.",
   "keywords": [
     "cli",