@aion0/forge 0.8.0 → 0.8.1

package/RELEASE_NOTES.md

@@ -1,107 +1,12 @@
-# Forge v0.8.0
+# Forge v0.8.1
 
 Released: 2026-05-19
 
-## Changes since v0.6.3
-
-### Features
-- feat: job preview auto-fill + forge worktree CLI + faster startup
-
-### Documentation
-- docs: example pipeline for a Forge Job driving Mantis → branch triage
-- docs: Jobs — scheduled connector polls that fan out to pipelines or chat
-- docs: connector declarative-extract spec + handoff + updated help doc
-- docs: connector packaging section + plugin category ADR
-- docs: ADR for browser-agent data architecture (Temper write strategy)
-- docs: RFC + implementation plan for browser-agent / connector system
+## Changes since v0.8.0
 
 ### Other
-- feat(chat): local memory fallback + simplified /chat web page
-- feat(security): encrypt plugin secrets at rest + mask on API
-- feat(pipeline): print fetch-bug-details + download-attachments to task log
-- feat(mantis+pipeline): attachment download into worktree for vision analysis
-- fix(pipeline): push-and-mr detects revoked glab token + prints fix cmd
-- fix(pipeline): push-and-mr surfaces glab output + MR-list fallback
-- feat(pipeline/mantis): fetch-bug-details exposes ALL fields downstream
-- fix(pipeline-ui): fetch full pipeline on click — summary breaks detail view
-- feat(jobs): expose {{total_matching}} placeholder in chat summary mode
-- fix(mantis): regex escape — YAML | block passes backslashes verbatim
-- fix(time): normalize all SQLite datetimes to ISO Z + add settings.timezone
-- feat(jobs/mantis): target='main' chat dispatch + total_matching count
-- fix(mantis): reset all filter fields before applying — no session leak
-- feat(chat): persistent main session + clear-messages + fork
-- feat(jobs): chat 'summary' dispatch mode — bundle N items into 1 LLM analysis
-- perf(pipeline-ui): lazy-load runs per workflow, drop 5s full poll
-- perf(watcher): defer initial sync + skip entry_count on first pass
-- feat(pipeline-ui): 'Load older runs' button per workflow
-- perf(pipeline): default limit 100 + ?workflow / ?before cursor
-- perf(pipeline): mtime cache + summary mode for list — 1.6MB → <50KB
-- perf(init): instrument 5s gap + demote skills sync failure to warn
-- feat(teams): _probe diagnostic tool — DOM state in one call
-- fix(chat): unwrap Node fetch 'fetch failed' with concrete remediation
-- fix(teams): char-by-char typing for search input (keyboard events)
-- feat(teams): v0.8.0 — send_message 3-tier resolver (left-rail → expand groups → search)
-- feat(mantis): v0.4.3 — accurate filters via show_status + handler_id
-- fix(pipeline): real workdir support + worktree-setup emits path only
-- fix(pipeline): node tasks always start fresh Claude session
-- fix(jobs): warn loudly when input_template keys render to empty
-- feat(jobs): 'Force run' — reset dedup + run, one click
-- fix(jobs): drop dedup_key from triggerPipeline — UNIQUE blocked replay
-- feat(gitlab): full HTTP-API connector (32 tools) — replaces DOM scraping
-- fix(mantis): bug notes always carry an id — fallback strategies + hash
-- feat(chat): connector narrowing — /teams forces tool use; bare mention narrows tool list
-- fix(chat): connector tools always surfaced + stronger 'use the tool' prompt
-- fix(notify-test): explain 'fetch failed' (corp SSL / proxy / DNS)
-- fix(notify-test): drop Markdown parse_mode + split chat_id + surface real reason
-- fix(mantis): search_bugs — fix_schedule (version) filter + auto-pagination
-- fix(mantis): search_bugs honours status + project_id, drops broken query syntax
-- feat(health): startup probe + Pipelines banner for missing CLIs (glab/gh/git/jq)
-- feat(pipeline): mantis-bug-fix-and-mr builtin + connector-tool API
-- feat(mantis): notes carry stable note_id so Jobs can dedup on them
-- feat(jobs): per-run verbose execution log persisted to job_runs.log
-- feat(web): Jobs first in Automation, deep-link to logs filtered by job id
-- fix(web): Automation sub-tabs squished header; new sub-toolbar row fix(terminal): Split Right/Down → SVG icons; mouse hint to tooltip feat(jobs): notes column — backfill + items_path mismatch surfaced
-- feat(web): Automation hub — merge Tasks/Pipelines nav + add Jobs view
-- feat(jobs): trace flow — Dashboard deep-link + CLI dispatches view
-- feat(jobs): Jobs backend — scheduled connector polls + dispatch
-- fix(settings): don't overwrite agent apiKey with masked placeholder
-- feat(chat): port Temper memory to Forge — pinned blocks + memory_* tools
-- feat(chat): reuse API agent profiles + add baseUrl (LiteLLM support)
-- feat(bridge): drop pairing — bridge auths with the user's Forge token
-- feat(connectors): Phase 3 — multi-protocol runtime (http, shell)
-- feat(chat): Phase 5 part 3 — Telegram /chat command
-- feat(chat): Phase 5 part 2 — forge chat CLI subcommand
-- feat(chat): Phase 5 part 1 — chat-standalone (8408) + Next proxy
-- fix(bridge): spawn browser-bridge from lib/init.ts too
-- feat(chat): Phase 2 part 3 — HTTP API + WS push streaming
-- feat(chat): Phase 2 part 2 — tool dispatcher + agent loop
-- feat(chat): Phase 2 part 1 — chat backend foundation
-- feat(bridge): Phase 1 — Forge↔Extension WebSocket bridge
-- feat(connectors): teams v0.7.1 — send_message no longer requires confirmation
-- feat(connectors): teams v0.7.0 — robust wait + send_message by name
-- feat(connectors): teams v0.6.0 — read_chat + read_channel by name
-- feat(connectors): teams v0.5.0 — channels support (list + read posts)
-- feat(connectors): pmdb v0.2.0 — read-only Fortinet PMDB connector
-- feat(connectors): teams v0.4.0 — send_message (destructive)
-- feat(connectors): teams v0.3.0 — rewritten from real DOM probe
-- feat(connectors): multi-runner — manifest declares execution context
-- feat(connectors): teams (Microsoft Teams) — first-cut declarative manifest
-- chore(connectors): bump gitlab to 0.2.0
-- feat(connectors): gitlab — declarative manifest with 6 tools
-- feat(connectors): declarative extract — manifest carries page + script
-- chore(connectors): bump mantis to 0.2.0 — full tool set landed
-- feat(mcp): user-managed external MCP servers in .mcp.json
-- feat(connectors): mantis probe — auto-copy JSON result to clipboard
-- fix(connectors): mantis probe — support 1.x classic URL + checkbox row marker
-- feat(connectors): mantis selector probe sibling script
-- chore(connectors): bump mantis + gitlab-browser to 0.1.1
-- docs(connectors): DOM extraction is the default path, not REST APIs
-- feat(ui): dedicated Connectors panel separate from Plugins
-- fix(connectors-api): rely on middleware auth, drop explicit token check
-- feat(api): connectors discovery + per-user settings endpoints
-- feat(plugins): connector category + browser-side mode, mantis + gitlab manifests
-- feat(gitlab): \`gitlab-issue-fix-and-review\` pipeline + scanner
-- fix(mcp): mirror .forge/mcp.json to project root so claude picks it up
+- feat(monitor): show Chat Server + Browser Bridge in Monitor panel
+- fix(chat): use relative imports in chat-standalone graph
 
 
-**Full Changelog**: https://github.com/aiwatching/forge/compare/v0.6.3...v0.8.0
+**Full Changelog**: https://github.com/aiwatching/forge/compare/v0.8.0...v0.8.1

package/app/api/monitor/route.ts

@@ -22,8 +22,25 @@ export async function GET() {
   const terminal = countProcess('terminal-standalone');
   const telegram = countProcess('telegram-standalone');
   const workspace = countProcess('workspace-standalone');
+  const chat = countProcess('chat-standalone');
+  const browserBridge = countProcess('browser-bridge-standalone');
   const tunnel = countProcess('cloudflared tunnel');
 
+  // Chat backend health (port 8408 — process can be alive but crashed)
+  let chatStatus: { running: boolean; sessions: number; port: number } = {
+    running: false,
+    sessions: 0,
+    port: Number(process.env.CHAT_PORT) || 8408,
+  };
+  try {
+    const chatRes = run(`curl -s http://127.0.0.1:${chatStatus.port}/api/status 2>/dev/null`);
+    if (chatRes) {
+      const data = JSON.parse(chatRes);
+      chatStatus.running = true;
+      chatStatus.sessions = data.active_sse_sessions ?? 0;
+    }
+  } catch {}
+
   // MCP Server (runs inside workspace process, check /health endpoint)
   let mcpStatus = { running: false, sessions: 0 };
   try {
@@ -65,6 +82,15 @@ export async function GET() {
       terminal: { running: terminal.count > 0, pid: terminal.pid, startedAt: terminal.startedAt },
       telegram: { running: telegram.count > 0, pid: telegram.pid, startedAt: telegram.startedAt },
       workspace: { running: workspace.count > 0, pid: workspace.pid, startedAt: workspace.startedAt },
+      browserBridge: { running: browserBridge.count > 0, pid: browserBridge.pid, startedAt: browserBridge.startedAt },
+      chat: {
+        running: chatStatus.running,
+        pid: chat.pid,
+        startedAt: chat.startedAt,
+        port: chatStatus.port,
+        sessions: chatStatus.sessions,
+        processAlive: chat.count > 0,
+      },
       tunnel: { running: tunnel.count > 0, pid: tunnel.pid, url: tunnelUrl, startedAt: tunnel.startedAt },
       mcp: { running: mcpStatus.running, port: 8406, sessions: mcpStatus.sessions },
     },

package/components/MonitorPanel.tsx

@@ -8,6 +8,15 @@ interface MonitorData {
     terminal: { running: boolean; pid: string; startedAt?: string };
     telegram: { running: boolean; pid: string; startedAt?: string };
     workspace: { running: boolean; pid: string; startedAt?: string };
+    browserBridge?: { running: boolean; pid: string; startedAt?: string };
+    chat?: {
+      running: boolean;
+      pid: string;
+      startedAt?: string;
+      port: number;
+      sessions: number;
+      processAlive: boolean;
+    };
     tunnel: { running: boolean; pid: string; url: string };
     mcp: { running: boolean; port: number; sessions: number };
   };
@@ -50,6 +59,7 @@ export default function MonitorPanel({ onClose }: { onClose: () => void }) {
                   { label: 'Terminal Server', ...data.processes.terminal },
                   { label: 'Telegram Bot', ...data.processes.telegram },
                   { label: 'Workspace Daemon', ...data.processes.workspace },
+                  ...(data.processes.browserBridge ? [{ label: 'Browser Bridge', ...data.processes.browserBridge }] : []),
                   { label: 'Tunnel', ...data.processes.tunnel },
                 ].map(p => (
                   <div key={p.label} className="flex items-center gap-2 text-xs">
@@ -65,6 +75,39 @@ export default function MonitorPanel({ onClose }: { onClose: () => void }) {
                     )}
                   </div>
                 ))}
+                {/* Chat Server (port 8408) — running = HTTP responds. If the
+                    process is alive but port isn't, the daemon likely crashed
+                    at boot (e.g. import error). */}
+                {data.processes.chat && (
+                  <div className="flex items-center gap-2 text-xs">
+                    <span
+                      className={
+                        data.processes.chat.running
+                          ? 'text-green-400'
+                          : data.processes.chat.processAlive
+                            ? 'text-yellow-500'
+                            : 'text-gray-500'
+                      }
+                    >
+                      ●
+                    </span>
+                    <span className="text-[var(--text-primary)] w-28">Chat Server</span>
+                    {data.processes.chat.running ? (
+                      <>
+                        <span className="text-[var(--text-secondary)] font-mono text-[10px]">port: {data.processes.chat.port}</span>
+                        <span className="text-gray-500 font-mono text-[9px]">{data.processes.chat.sessions} SSE</span>
+                        {data.processes.chat.pid && (
+                          <span className="text-[var(--text-secondary)] font-mono text-[10px]">pid: {data.processes.chat.pid}</span>
+                        )}
+                      </>
+                    ) : data.processes.chat.processAlive ? (
+                      <span className="text-yellow-500 text-[10px]">port {data.processes.chat.port} unreachable (check forge.log)</span>
+                    ) : (
+                      <span className="text-gray-500 text-[10px]">stopped</span>
+                    )}
+                  </div>
+                )}
+
                 {/* MCP Server — separate row (no pid, uses port + sessions) */}
                 <div className="flex items-center gap-2 text-xs">
                   <span className={data.processes.mcp?.running ? 'text-green-400' : 'text-gray-500'}>●</span>

package/lib/chat/agent-loop.ts

@@ -12,7 +12,7 @@
  */
 
 import { streamLlm, type LlmTool } from './llm';
-import { loadSettings } from '@/lib/settings';
+import { loadSettings } from '../settings';
 import {
   appendMessage,
   getSession,
@@ -31,7 +31,7 @@ import {
   getPlugin,
   getInstalledPlugin,
   getConnectorsForPlugin,
-} from '@/lib/plugins/registry';
+} from '../plugins/registry';
 import type {
   ContentBlock,
   Message,

package/lib/chat/local-memory.ts

@@ -17,8 +17,8 @@
  */
 
 import { join } from 'node:path';
-import { getDb } from '@/src/core/db/database';
-import { getDataDir } from '@/lib/dirs';
+import { getDb } from '../../src/core/db/database';
+import { getDataDir } from '../dirs';
 import type Database from 'better-sqlite3';
 import type {
   EpisodeInput,

package/lib/chat/memory-store.ts

@@ -15,7 +15,7 @@
  * so the chat loop registers the memory_* tools either way.
  */
 
-import { loadSettings } from '@/lib/settings';
+import { loadSettings } from '../settings';
 import { TemperClient } from './temper';
 import { LocalMemoryStore } from './local-memory';
 import type { EpisodeInput, MemoryBlock, SearchHit } from './temper';

package/lib/chat/protocols/http.ts

@@ -15,8 +15,8 @@
  * is_error so the LLM can react.
  */
 
-import type { HttpRequestSpec, ConnectorTool } from '@/lib/plugins/types';
-import { expandAllTokens } from '@/lib/plugins/templates';
+import type { HttpRequestSpec, ConnectorTool } from '../../plugins/types';
+import { expandAllTokens } from '../../plugins/templates';
 
 export interface HttpProtocolArgs {
   tool: ConnectorTool;

package/lib/chat/protocols/shell.ts

@@ -15,8 +15,8 @@
  * automatic command allow-list in v1.
  */
 
-import type { ConnectorTool } from '@/lib/plugins/types';
-import { expandAllTokens } from '@/lib/plugins/templates';
+import type { ConnectorTool } from '../../plugins/types';
+import { expandAllTokens } from '../../plugins/templates';
 import { spawn } from 'node:child_process';
 
 export interface ShellProtocolArgs {

package/lib/chat/session-store.ts

@@ -7,8 +7,8 @@
 
 import { randomUUID } from 'node:crypto';
 import { join } from 'node:path';
-import { getDb } from '@/src/core/db/database';
-import { getDataDir } from '@/lib/dirs';
+import { getDb } from '../../src/core/db/database';
+import { getDataDir } from '../dirs';
 import type { ContentBlock, Message, Role, Session } from './types';
 
 function db() {

package/lib/chat/tool-dispatcher.ts

@@ -16,9 +16,9 @@ import {
   getPlugin,
   getInstalledPlugin,
   getConnectorsForPlugin,
-} from '@/lib/plugins/registry';
-import { expandSettingsTokens } from '@/lib/plugins/templates';
-import type { Connector, ConnectorTool } from '@/lib/plugins/types';
+} from '../plugins/registry';
+import { expandSettingsTokens } from '../plugins/templates';
+import type { Connector, ConnectorTool } from '../plugins/types';
 
 export interface ToolCall {
   id: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aion0/forge",
-  "version": "0.8.0",
+  "version": "0.8.1",
   "description": "Unified AI workflow platform — multi-model task orchestration, persistent sessions, web terminal, remote access",
   "type": "module",
   "scripts": {