@aion0/forge 0.10.17 → 0.10.18

package/RELEASE_NOTES.md

@@ -1,11 +1,8 @@
-# Forge v0.10.17
+# Forge v0.10.18
 
 Released: 2026-05-30
 
-## Changes since v0.10.16
+## Changes since v0.10.17
 
-### Features
-- feat: API profile model picker uses registry providers
 
-
-**Full Changelog**: https://github.com/aiwatching/forge/compare/v0.10.16...v0.10.17
+**Full Changelog**: https://github.com/aiwatching/forge/compare/v0.10.17...v0.10.18

package/components/ProjectDetail.tsx

@@ -1678,7 +1678,7 @@ function AgentTerminalButton({ projectPath, projectName }: { projectPath: string
         currentSessionId,
       });
     } catch {
-      openTerminal(a.id);
+      openTerminal(a.id, false);
     }
   };
 

package/components/WebTerminal.tsx

@@ -398,9 +398,11 @@ const WebTerminal = forwardRef<WebTerminalHandle, WebTerminalProps>(function Web
         agentSkipFlag = agentConfig?.skipPermissionsFlag || '';
       } catch {}
 
-      // Resume flag: explicit sessionId > fixedSession > -c (only for session-capable agents)
+      // Resume flag: explicit sessionId > fixedSession > -c (only for session-capable agents).
+      // resumeMode === false means the user explicitly chose "New Session" — never inject
+      // a fixedSession then, or "new" silently resumes the current one.
       let resumeFlag = '';
-      if (supportsSession) {
+      if (supportsSession && resumeMode !== false) {
         if (sessionId) resumeFlag = ` --resume ${sessionId}`;
         else if (resumeMode) resumeFlag = ' -c';
         // Override with fixedSession if no explicit sessionId
@@ -987,9 +989,12 @@ const WebTerminal = forwardRef<WebTerminalHandle, WebTerminalProps>(function Web
                                         currentSessionId,
                                       });
                                     } catch {
-                                      // Fallback: open directly without picker
+                                      // Fallback: open directly without picker.
+                                      // resumeMode:false so this fresh-open path skips the
+                                      // fixedSession resume branch downstream (else "open"
+                                      // silently resumes the current session).
                                       window.dispatchEvent(new CustomEvent('forge:open-terminal', {
-                                        detail: { projectPath: p.path, projectName: p.name, agentId: a.id },
+                                        detail: { projectPath: p.path, projectName: p.name, agentId: a.id, resumeMode: false },
                                       }));
                                     }
                                   }}
@@ -1068,7 +1073,10 @@ const WebTerminal = forwardRef<WebTerminalHandle, WebTerminalProps>(function Web
           onSelect={(sel) => {
             const info = vibePickerInfo;
             setVibePickerInfo(null);
-            const detail: any = { projectPath: info.projectPath, projectName: info.projectName, agentId: info.agentId, profileEnv: info.profileEnv };
+            // resumeMode must be LITERAL false for "new" — the downstream guard
+            // is `resumeMode !== false`, so leaving it undefined lets fixedSession
+            // get injected and "New session" silently resumes the current one.
+            const detail: any = { projectPath: info.projectPath, projectName: info.projectName, agentId: info.agentId, profileEnv: info.profileEnv, resumeMode: false };
             if (sel.mode !== 'new') { detail.resumeMode = true; detail.sessionId = sel.sessionId; }
             window.dispatchEvent(new CustomEvent('forge:open-terminal', { detail }));
           }}

package/lib/agents/index.ts

@@ -280,7 +280,12 @@ export function resolveTerminalLaunch(agentId?: string): TerminalLaunchInfo {
   }
 
   return {
-    cliCmd: cli.cmd,
+    // Prefer the user-configured absolute binary path over the bare command
+    // name. A bare `claude` resolves via the tmux pane's PATH — non-login shell
+    // + forge-server's inherited PATH can pick a different/older install than
+    // the user's interactive shell, and an old claude crashes resuming a
+    // session a newer claude wrote. An absolute path bypasses PATH entirely.
+    cliCmd: agentCfg.path || cli.cmd,
     cliType,
     supportsSession: cli.session,
     resumeFlag: agentCfg.resumeFlag || cli.resume,

package/lib/chat/tool-dispatcher.ts

@@ -189,6 +189,25 @@ const BUILTINS: Record<string, BuiltinHandler> = {
     return line;
   },
 
+  // Query a pipeline run's status + per-node results by id (pairs with
+  // trigger_pipeline's returned id). Mirrors the MCP get_pipeline_status tool.
+  get_pipeline_status: async (input) => {
+    const params = (input as { pipeline_id?: string } | undefined) || {};
+    if (!params.pipeline_id) return 'get_pipeline_status failed: pipeline_id is required (returned by trigger_pipeline).';
+    const { getPipeline } = await import('../pipeline');
+    const pipeline = getPipeline(params.pipeline_id);
+    if (!pipeline) return `Pipeline "${params.pipeline_id}" not found.`;
+    const nodes = Object.entries(pipeline.nodes || {}).map(([id, n]) => {
+      let line = `  ${id}: ${n.status}`;
+      if (n.error) line += ` — ${n.error}`;
+      for (const [k, v] of Object.entries(n.outputs || {})) {
+        line += `\n    ${k}: ${String(v).slice(0, 200)}`;
+      }
+      return line;
+    }).join('\n');
+    return `Pipeline ${pipeline.id} [${pipeline.status}] (${pipeline.workflowName})\n${nodes}`;
+  },
+
   // Surface Forge's local context (projects + agents + skills) so the chat
   // agent can pick valid values for inputs like trigger_pipeline.input.project
   // without guessing. Cheap call — read-only directory + DB lookups.
@@ -291,6 +310,20 @@ export const BUILTIN_TOOL_DEFS: BuiltinToolDef[] = [
       },
     },
   },
+  {
+    name: 'get_pipeline_status',
+    description: "Check a Forge pipeline run's live status + per-node results by id. Pass pipeline_id (returned by trigger_pipeline) to get the run's overall status + each node's status / error / outputs. Use whenever the user asks how a running or finished pipeline is doing.",
+    input_schema: {
+      type: 'object',
+      properties: {
+        pipeline_id: {
+          type: 'string',
+          description: 'Pipeline run id (returned by trigger_pipeline).',
+        },
+      },
+      required: ['pipeline_id'],
+    },
+  },
   {
     name: 'list_forge_context',
     description: "Return the current Forge instance's local context: project names (use these as input.project for pipelines / dispatch_task), agent profile ids, and installed skills. Call this whenever the user references a project / agent / skill by name and you need to validate the name OR when picking defaults for trigger_pipeline / dispatch_task. No arguments.",

package/lib/help-docs/05-pipelines.md

@@ -639,6 +639,15 @@ stop_condition: "all agents say DONE"
 initial_prompt: "Review PR #{{input.pr_number}} in project {{input.project}}. Reviewer: analyze the diff. Author: be ready to address feedback."
 ```
 
+## Controlling pipelines from chat
+
+The chat assistant has builtin tools for pipelines, so you can run and check them in plain language:
+
+- **`trigger_pipeline`** — start a workflow. Ask the assistant to "run the `<name>` pipeline" and it dispatches the run, returning a pipeline run id. Call it with no arguments first and it lists every workflow plus its input schema.
+- **`get_pipeline_status`** — check a run by id. Ask "what's the status of pipeline `<id>`?" and it returns the run's overall status plus each node's status, error, and outputs.
+
+Typical flow: trigger a pipeline, note the returned id, then ask the assistant for its status whenever you want an update — no need to open the Pipelines view. The same two tools are exposed to Claude Code over MCP (`trigger_pipeline` / `get_pipeline_status`).
+
 ## Pipeline Model
 
 In **Settings → Pipeline Model**, you can select which Claude model runs pipeline tasks. Set to `default` to use the same model as regular tasks.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aion0/forge",
-  "version": "0.10.17",
+  "version": "0.10.18",
   "description": "Unified AI workflow platform — multi-model task orchestration, persistent sessions, web terminal, remote access",
   "type": "module",
   "scripts": {