diff options
Diffstat (limited to 'modules')
| -rw-r--r-- | modules/core/core-bundle.js | 200 | ||||
| -rw-r--r-- | modules/core/module-aiassist.js | 62 | ||||
| -rw-r--r-- | modules/core/module-billing.js | 33 | ||||
| -rw-r--r-- | modules/core/module-fetcher.js | 17 | ||||
| -rw-r--r-- | modules/core/module-loader.js | 2 | ||||
| -rw-r--r-- | modules/core/module-orchestrator.js | 88 | ||||
| -rw-r--r-- | modules/ext/cryptogzip.js | 67 | ||||
| -rw-r--r-- | modules/ext/media.js | 6 | ||||
| -rw-r--r-- | modules/ext/webdav.js | 20 | ||||
| -rw-r--r-- | modules/test/test-orchestrator-parallel-harvest.js | 132 |
10 files changed, 516 insertions, 111 deletions
diff --git a/modules/core/core-bundle.js b/modules/core/core-bundle.js index 8b6e4cb..4483709 100644 --- a/modules/core/core-bundle.js +++ b/modules/core/core-bundle.js @@ -7773,8 +7773,10 @@ class OllamaModule { // Billing / pricing surface for newOS4 (2026-10-28) // // Why: the terminal could show a model's name but never what it costs. Prices and the anonymous -// message allowance are public data, so they need no login — the account BALANCE does, and newOS4 -// has no session layer, so that is reported honestly as "not signed in" rather than guessed. +// message allowance are public data, so they need no login — the account BALANCE does. +// 2026-11-03: newOS4 DOES have a session — the browser sends the HttpOnly cookie and +// parse_bearer_token() reads it — so the balance below is real when logged in. The old comment +// claimed otherwise and made the 401 look expected. // // Commands: cost [on|off], usage class BillingModule { @@ -7828,7 +7830,9 @@ class BillingModule { } async _loadBalance() { - // Only answers when a session exists; newOS4 currently has none, so this is expected to 401. + // Cookie auth is what makes this work: the browser sends webx11_session_token and + // parse_bearer_token() falls back to it. If this 401s, the user is logged out — say so, + // do not assume newOS4 cannot authenticate (that comment was wrong until 2026-11-03). try { const r = await fetch('/api/billing/subscription', { credentials: 'include' }); if (!r.ok) return null; @@ -7839,6 +7843,16 @@ class BillingModule { } } + async _loadUsage(days) { + try { + const r = await fetch('/api/billing/usage?days=' + (days || 1), { credentials: 'include' }); + if (!r.ok) return null; + return await r.json(); + } catch (e) { + return null; + } + } + // --- session spend, measured where the tokens actually are ---------------- _hookUsage() { const om = this.terminal && this.terminal.ollamaManager; @@ -7881,8 +7895,10 @@ class BillingModule { } } + // On by default since 2026-11-03: "the terminal shows tokens but not cost" was the default, + // not a bug — the flag was opt-in and nobody knew the command existed. `cost off` opts out. _costLineOn() { - try { return localStorage.getItem('webx11_cost_line') === '1'; } catch (e) { return false; } + try { return localStorage.getItem('webx11_cost_line') !== '0'; } catch (e) { return true; } } // --- the command ---------------------------------------------------------- @@ -7910,13 +7926,20 @@ class BillingModule { p('📊 Free allowance: unavailable (billing API did not answer)', 'warning'); } - p(`💰 Balance: ${bal && bal.balance_usd != null ? '$' + Number(bal.balance_usd).toFixed(2) : 'not signed in — newOS4 sends no session token, so a balance cannot be shown here yet'}`, 'info'); + p(`💰 Balance: ${bal && bal.balance_usd != null ? '$' + Number(bal.balance_usd).toFixed(2) : 'not signed in — log in at /spaces/dashboard.html to see your balance'}`, 'info'); if (bal && bal.plan_allowance) { const pa = bal.plan_allowance; p(`🎁 Included usage (${pa.plan}): $${Number(pa.used_usd).toFixed(2)} of $${Number(pa.limit_usd).toFixed(2)} ${pa.period === 'day' ? 'today' : 'this month'} — $${Number(pa.remaining_usd).toFixed(2)} left, then your balance pays`, 'info'); } + const u = await this._loadUsage(1); + if (u && u.totals) { + p(`📅 Today (all newOS4 sessions): ${u.totals.requests} call(s), ` + + `${((u.totals.prompt_tokens + u.totals.completion_tokens) / 1000).toFixed(1)}k tokens, ` + + `≈ $${Number(u.totals.retail_usd).toFixed(2)} retail`, 'info'); + } + if (s && s.msgs) { p(`🧮 This session: ${s.msgs} call(s), ${(s.prompt / 1000).toFixed(1)}k in / ${(s.completion / 1000).toFixed(1)}k out ≈ $${s.cost.toFixed(4)} retail`, 'info'); Object.entries(s.byModel) @@ -9121,7 +9144,8 @@ buildSystemPrompt() { method: { type: 'string', enum: ['GET', 'POST'], description: 'HTTP method (default GET)' }, body: { type: 'string', description: 'Request body for POST (JSON string, max 10KB)' }, content_type: { type: 'string', description: 'Content-Type header (default application/json for POST)' }, -max_chars: { type: 'integer', description: 'Max characters of page content to return (default from the model profile; max 50000)' }, +max_chars: { type: 'integer', description: 'Max characters of page content to return (default from the model profile; a page is capped by the tool-result budget, so page with offset/next_offset for longer pages)' }, + offset: { type: 'integer', description: 'Start character offset into the extracted text (default 0). Use the next_offset from the previous reply to continue.' }, }, required: ['url'] } @@ -10350,6 +10374,11 @@ max_chars: { type: 'integer', description: 'Max characters of page content to re if (keep < obj.content.length) { const c = obj.content.charCodeAt(keep - 1); if (c >= 0xD800 && c <= 0xDBFF) keep -= 1; + } else if (obj.content.length <= keep) { + // The page fits; only the envelope is fat. Do not claim a loss that + // did not happen (measured 2026-11-06: "clipped: 0 of 11800 withheld" + // appended while every character was returned). + return JSON.stringify(obj); } const hadMore = (obj.next_offset !== null && obj.next_offset !== undefined); const full = obj.content.length; @@ -11601,9 +11630,14 @@ max_chars: { type: 'integer', description: 'Max characters of page content to re const contentType = a.content_type || (method === 'POST' ? 'application/json' : ''); try { // Use /api/fetch for POST, /api/extract for GET (extract is cleaner for readable content) - const maxChars = Math.max(500, Math.min(parseInt(a.max_chars, 10) || this._toolLimit('fetchChars'), 50000)); + // A page larger than toolResultChars-200 is a page the model never sees + // (measured 2026-11-06: asked 50 000, received 11 800, no next_offset). + const wanted = parseInt(a.max_chars, 10) || this._toolLimit('fetchChars'); + const clipMax = this._toolLimit('toolResultChars'); + const limit = Math.max(500, Math.min(wanted, clipMax - 200)); // request window + const offset = Math.max(0, parseInt(a.offset, 10) || 0); const endpoint = method === 'POST' ? '/api/fetch' : '/api/extract'; - const params = new URLSearchParams({ url, max_chars: String(maxChars) }); + const params = new URLSearchParams({ url, max_chars: String(Math.min(100000, offset + limit)) }); if (method === 'POST') { params.set('method', 'POST'); params.set('body', body); if (contentType) params.set('content_type', contentType); } const resp = await fetch(endpoint + '?' + params.toString()); if (resp.status === 403) return JSON.stringify({ error: 'URL blocked — internal network' }); @@ -11611,18 +11645,47 @@ max_chars: { type: 'integer', description: 'Max characters of page content to re const err = await resp.json().catch(() => ({})); return JSON.stringify({ error: 'Fetch failed: ' + (err.error || `HTTP ${resp.status}`), status: resp.status, details: err.details || '' }); } - // For POST/route, the response is raw text; for GET/extract it's JSON + // POST: /api/fetch returns raw text, so page it by characters. if (method === 'POST') { const text = await resp.text(); - return JSON.stringify({ method, content: text.substring(0, maxChars), content_chars: text.length, - returned_chars: Math.min(text.length, maxChars), truncated: text.length > maxChars }); + const start = Math.min(offset, text.length); + let lim = Math.max(500, Math.min(limit, clipMax - 200)); + let env; + do { + const slice = text.slice(start, start + lim); + const nextOffset = start + slice.length < text.length ? start + slice.length : null; + env = { method, offset: start, content: slice, returned_chars: slice.length, + content_chars: text.length, total_chars: text.length, source_bytes: text.length, + truncated: nextOffset !== null, next_offset: nextOffset }; + if (JSON.stringify(env).length <= clipMax || lim <= 500) break; + lim -= Math.max(1, JSON.stringify(env).length - clipMax); + } while (true); + return JSON.stringify(env); } + // GET: /api/extract carries content_chars = the true length of the extracted + // text, so offset/next_offset are exact rather than hopeful. const data = await resp.json(); - return JSON.stringify({ method, title: data.title || '', content: (data.content || '').substring(0, maxChars), - content_chars: data.content_chars || (data.content || '').length, - returned_chars: Math.min((data.content || '').length, maxChars), - truncated: !!(data.truncated || (data.content_chars > maxChars)), - word_count: data.word_count || 0, method_proxy: data.method || 'extract' }); + const full = data.content || ''; + // UTF-16 length when the proxy supplies it: Python len() counts code points, + // JS counts UTF-16 units, and 7 astral chars were enough to leave a 7-char + // gap at the end of a 90 KB page (measured 2026-11-06). + const totalChars = Number(data.content_chars_utf16) || Number(data.content_chars) || full.length; + let start = Math.min(offset, totalChars); + if (start > 0) { const prev = full.charCodeAt(start - 1); if (prev >= 0xD800 && prev <= 0xDBFF) start += 1; } + let lim = Math.max(500, Math.min(limit, clipMax - 200)); + let env; + do { + const slice = full.slice(start, start + lim); + const nextOffset = start + slice.length < totalChars ? start + slice.length : null; + env = { method, title: data.title || '', offset: start, content: slice, returned_chars: slice.length, + content_chars: totalChars, total_chars: totalChars, + source_bytes: (data.source_bytes === undefined ? null : data.source_bytes), + truncated: nextOffset !== null, next_offset: nextOffset, + word_count: data.word_count || 0, method_proxy: data.method || 'extract' }; + if (JSON.stringify(env).length <= clipMax || lim <= 500) break; + lim -= Math.max(1, JSON.stringify(env).length - clipMax); + } while (true); + return JSON.stringify(env); } catch (e) { return JSON.stringify({ error: 'Fetch failed: ' + e.message }); } } case 'wpm_install': { @@ -14039,7 +14102,7 @@ class WebFetcherModule { } getHelpInfo() { return { - 'fetch': 'fetch <url> [--post=BODY] [--type=CT] [--raw] — Fetch URL (raw HTML/text); --post enables POST with body, --type sets Content-Type, --raw shows untruncated content', + 'fetch': 'fetch <url> [--post=BODY] [--type=CT] [--raw] [--save=FILE] — Fetch URL (raw HTML/text); --post enables POST with body, --type sets Content-Type, --raw shows untruncated content, --save writes it to the VFS', 'extract': 'extract <url> — Extract readable article content (title + body text) from a URL', }; } @@ -14103,6 +14166,21 @@ class WebFetcherModule { return; } const text = await resp.text(); + // --save: the only terminal path from a remote URL to the VFS. Without it a + // 100 KB source page can only be read through 3 000-char scrollback, which is + // what sent the in-OS agent looking for a shell redirect (reported 2026-11-06). + if (flags.save) { + const path = String(flags.save === true ? (url.split('/').pop().split('?')[0] || 'fetch.txt') : flags.save); + const dir = path.includes('/') ? path.slice(0, path.lastIndexOf('/')) : ''; + if (dir) { try { this.terminal.fs.mkdir(dir); } catch (e) {} } + const wrote = this.terminal.fs.writeFile(path, text); + if (!wrote || this.terminal.fs.readFile(path) !== text) { + this.terminal.print(`Save failed: could not write ${path}`, 'error'); + return; + } + this.terminal.print(`💾 Saved ${text.length} chars to ${path}`, 'success'); + return; + } const limit = showRaw ? text.length : 3000; this.terminal.print(`\n=== Content from ${url} (${method}) ===\n`, 'success'); this.terminal.print(text.substring(0, limit) + (text.length > limit && !showRaw ? '...' : '')); @@ -17464,6 +17542,18 @@ class OrchestratorModule { if (!task) return this.terminal.print('Usage: orchestrate <task description>', 'error'); this._delegations = (this._delegations || 0) + 1; + // Parallel fan-out bookkeeping. A sub-task carries its own id, and the batch that + // spawned it hands in a sink, so the harvest below matches results by identity — + // never by a task string that a later delegation on the same agent can overwrite. + const taskId = opts.taskId || null; + const resultSink = opts.resultSink || null; + const recordResult = (status, payload) => { + if (!taskId || !resultSink || typeof resultSink.set !== 'function') return; + resultSink.set(taskId, Object.assign({ + taskId, task, status, at: new Date().toISOString() + }, payload || {})); + }; + // Parallel fan-out: decompose and run sub-tasks concurrently if (opts.parallel) { const parts = this._decomposeGoal(task); @@ -17484,38 +17574,45 @@ class OrchestratorModule { } } - const results = await Promise.allSettled(parts.map(sub => this.delegateTask(sub.goal))); - const successes = results.filter(r => r.status === 'fulfilled').length; - const failures = results.filter(r => r.status === 'rejected').length; + const runId = 'pb' + Date.now().toString(36) + '-' + (this._parallelSeq = (this._parallelSeq || 0) + 1); + const resultSink = new Map(); + const results = await Promise.allSettled(parts.map((sub, i) => + this.delegateTask(sub.goal, { taskId: runId + ':' + i, resultSink }))); - // Result synthesis: collect outbox results from each sub-task's agent + // Result synthesis: one row per sub-task, looked up by the id handed to it. const summaries = []; for (let i = 0; i < parts.length; i++) { const sub = parts[i]; const r = results[i]; - if (r.status === 'fulfilled') { - // Find the agent that handled this sub-task - const agent = Object.values(this._agents).find(a => a.currentTask === sub.goal || (a.lastTask === sub.goal)); - if (!agent) { - // "ok / completed (agent recycled)" told the caller a task succeeded - // when we could not even find its agent. - summaries.push({ task: sub.goal, status: 'failed', error: 'agent not found — result unverifiable' }); - continue; - } - const outbox = this.terminal.fs.fsReadJSON(`/agents/${agent.id}/outbox.json`); - const body = outbox && typeof outbox.result === 'string' ? outbox.result : ''; - const snippet = body.substring(0, 200); + const subTaskId = runId + ':' + i; + const rec = resultSink.get(subTaskId); + if (rec) { summaries.push({ - task: sub.goal, agent: agent.id, - status: body ? 'ok' : 'failed', - result: body || null, - error: body ? undefined : (outbox && outbox.error) || 'no result in outbox' + task: sub.goal, + taskId: subTaskId, + agent: rec.agentId, + status: rec.status === 'ok' ? 'ok' : 'failed', + result: rec.status === 'ok' ? (rec.result || null) : null, + error: rec.status === 'ok' ? undefined : (rec.error || 'no result recorded') }); - } else { - summaries.push({ task: sub.goal, status: 'failed', error: r.reason?.message || 'unknown error' }); + continue; } + // The call died before it could report itself (a throw from a path that never + // reached a recordResult call): the promise state is the only evidence left. + summaries.push({ + task: sub.goal, + taskId: subTaskId, + status: 'failed', + error: (r.status === 'rejected' && ((r.reason && r.reason.message) || String(r.reason))) || 'no result recorded' + }); } + // The tally IS the summaries — never the promise states. allSettled counts + // resolutions, and a spawn refusal returns terminal.print() (undefined), which + // FULFILS: that is how "3/3 succeeded" printed over three failed rows. + const successes = summaries.filter(s => s.status === 'ok').length; + const failures = summaries.length - successes; + // Store merged result to VFS for persistence const merged = { type: 'parallel_result', @@ -17537,7 +17634,8 @@ class OrchestratorModule { const snippet = s.result ? s.result.replace(/\n/g, ' ').substring(0, 60) : ''; this.terminal.print(' ' + icon + ' ' + s.task.substring(0, 35) + agentTag + ' → ' + snippet, 'info'); } - this.terminal.print('✅ Parallel batch: ' + successes + '/' + parts.length + ' succeeded' + (failures > 0 ? ', ' + failures + ' failed' : ''), 'header'); + // The header must agree with the rows above it: ✅ only when nothing failed. + this.terminal.print((failures === 0 ? '✅' : '❌') + ' Parallel batch: ' + successes + '/' + parts.length + ' succeeded' + (failures > 0 ? ', ' + failures + ' failed' : ''), failures === 0 ? 'header' : 'error'); return merged; } // Single sub-task — fall through to normal path @@ -17556,8 +17654,10 @@ class OrchestratorModule { // Auto-spawn with role-derived capabilities if (!agent) { if (Object.keys(this._agents).length >= this._SPAWN_LIMIT) { - // Try finding any non-failed agent to reuse - const reuse = Object.values(this._agents).find(a => a.status !== 'failed' && a.status !== 'sleeping'); + // Reuse an agent that is NOT running. A BUSY agent must never be handed a + // second task: concurrent sub-tasks then overwrite currentTask/leaseId and + // erase each other's results (reproduced 4/4 — Defect 2b). + const reuse = Object.values(this._agents).find(a => a.status !== 'busy' && a.status !== 'failed'); if (reuse) { reuse.status = 'busy'; agent = reuse; @@ -17567,11 +17667,18 @@ class OrchestratorModule { const role = this._deriveRole(task); const caps = this._deriveCapabilities(task); agent = this.spawnAgent(role, caps); - if (!agent) return this.terminal.print('❌ No available agents — spawn limit reached', 'error'); + if (!agent) { + const msg = 'no idle agent — ' + Object.keys(this._agents).length + '/' + this._SPAWN_LIMIT + ' agents busy, refusing to share one'; + recordResult('failed', { error: msg }); + return this.terminal.print('❌ ' + msg, 'error'); + } agent.status = 'busy'; // claim immediately after spawn } } - if (!agent) return this.terminal.print('❌ No available agents', 'error'); + if (!agent) { + recordResult('failed', { error: 'no agent available' }); + return this.terminal.print('❌ No available agents', 'error'); + } // A2A Contract handshake const registry = this.terminal.moduleManager.modules['AgentRegistryModule']; @@ -17641,6 +17748,9 @@ class OrchestratorModule { // Return result for callers that await delegateTask outboxResult = _answer; // '' when the worker returned nothing; callers see that + recordResult(_answer ? 'ok' : 'failed', _answer + ? { result: _answer, agentId: agent.id } + : { agentId: agent.id, error: 'agent produced no answer' }); // Check restart policy — if always, set back to idle for next task if (agent.restartPolicy === 'always') { @@ -17652,8 +17762,10 @@ class OrchestratorModule { } else { this.terminal.print(`ℹ️ AI module not available — task queued for ${agent.id}`, 'info'); outboxResult = 'queued — AI module not loaded, nothing executed'; + recordResult('failed', { agentId: agent.id, error: 'AI module not loaded — nothing executed' }); } } catch (e) { + recordResult('failed', { agentId: agent.id, error: e.message || String(e) }); this._handleAgentFailure(agent.id, e.message); } diff --git a/modules/core/module-aiassist.js b/modules/core/module-aiassist.js index 0a0ea71..a3ffb6f 100644 --- a/modules/core/module-aiassist.js +++ b/modules/core/module-aiassist.js @@ -226,7 +226,8 @@ buildSystemPrompt() { method: { type: 'string', enum: ['GET', 'POST'], description: 'HTTP method (default GET)' }, body: { type: 'string', description: 'Request body for POST (JSON string, max 10KB)' }, content_type: { type: 'string', description: 'Content-Type header (default application/json for POST)' }, -max_chars: { type: 'integer', description: 'Max characters of page content to return (default from the model profile; max 50000)' }, +max_chars: { type: 'integer', description: 'Max characters of page content to return (default from the model profile; a page is capped by the tool-result budget, so page with offset/next_offset for longer pages)' }, + offset: { type: 'integer', description: 'Start character offset into the extracted text (default 0). Use the next_offset from the previous reply to continue.' }, }, required: ['url'] } @@ -1455,6 +1456,11 @@ max_chars: { type: 'integer', description: 'Max characters of page content to re if (keep < obj.content.length) { const c = obj.content.charCodeAt(keep - 1); if (c >= 0xD800 && c <= 0xDBFF) keep -= 1; + } else if (obj.content.length <= keep) { + // The page fits; only the envelope is fat. Do not claim a loss that + // did not happen (measured 2026-11-06: "clipped: 0 of 11800 withheld" + // appended while every character was returned). + return JSON.stringify(obj); } const hadMore = (obj.next_offset !== null && obj.next_offset !== undefined); const full = obj.content.length; @@ -2706,9 +2712,14 @@ max_chars: { type: 'integer', description: 'Max characters of page content to re const contentType = a.content_type || (method === 'POST' ? 'application/json' : ''); try { // Use /api/fetch for POST, /api/extract for GET (extract is cleaner for readable content) - const maxChars = Math.max(500, Math.min(parseInt(a.max_chars, 10) || this._toolLimit('fetchChars'), 50000)); + // A page larger than toolResultChars-200 is a page the model never sees + // (measured 2026-11-06: asked 50 000, received 11 800, no next_offset). + const wanted = parseInt(a.max_chars, 10) || this._toolLimit('fetchChars'); + const clipMax = this._toolLimit('toolResultChars'); + const limit = Math.max(500, Math.min(wanted, clipMax - 200)); // request window + const offset = Math.max(0, parseInt(a.offset, 10) || 0); const endpoint = method === 'POST' ? '/api/fetch' : '/api/extract'; - const params = new URLSearchParams({ url, max_chars: String(maxChars) }); + const params = new URLSearchParams({ url, max_chars: String(Math.min(100000, offset + limit)) }); if (method === 'POST') { params.set('method', 'POST'); params.set('body', body); if (contentType) params.set('content_type', contentType); } const resp = await fetch(endpoint + '?' + params.toString()); if (resp.status === 403) return JSON.stringify({ error: 'URL blocked — internal network' }); @@ -2716,18 +2727,47 @@ max_chars: { type: 'integer', description: 'Max characters of page content to re const err = await resp.json().catch(() => ({})); return JSON.stringify({ error: 'Fetch failed: ' + (err.error || `HTTP ${resp.status}`), status: resp.status, details: err.details || '' }); } - // For POST/route, the response is raw text; for GET/extract it's JSON + // POST: /api/fetch returns raw text, so page it by characters. if (method === 'POST') { const text = await resp.text(); - return JSON.stringify({ method, content: text.substring(0, maxChars), content_chars: text.length, - returned_chars: Math.min(text.length, maxChars), truncated: text.length > maxChars }); + const start = Math.min(offset, text.length); + let lim = Math.max(500, Math.min(limit, clipMax - 200)); + let env; + do { + const slice = text.slice(start, start + lim); + const nextOffset = start + slice.length < text.length ? start + slice.length : null; + env = { method, offset: start, content: slice, returned_chars: slice.length, + content_chars: text.length, total_chars: text.length, source_bytes: text.length, + truncated: nextOffset !== null, next_offset: nextOffset }; + if (JSON.stringify(env).length <= clipMax || lim <= 500) break; + lim -= Math.max(1, JSON.stringify(env).length - clipMax); + } while (true); + return JSON.stringify(env); } + // GET: /api/extract carries content_chars = the true length of the extracted + // text, so offset/next_offset are exact rather than hopeful. const data = await resp.json(); - return JSON.stringify({ method, title: data.title || '', content: (data.content || '').substring(0, maxChars), - content_chars: data.content_chars || (data.content || '').length, - returned_chars: Math.min((data.content || '').length, maxChars), - truncated: !!(data.truncated || (data.content_chars > maxChars)), - word_count: data.word_count || 0, method_proxy: data.method || 'extract' }); + const full = data.content || ''; + // UTF-16 length when the proxy supplies it: Python len() counts code points, + // JS counts UTF-16 units, and 7 astral chars were enough to leave a 7-char + // gap at the end of a 90 KB page (measured 2026-11-06). + const totalChars = Number(data.content_chars_utf16) || Number(data.content_chars) || full.length; + let start = Math.min(offset, totalChars); + if (start > 0) { const prev = full.charCodeAt(start - 1); if (prev >= 0xD800 && prev <= 0xDBFF) start += 1; } + let lim = Math.max(500, Math.min(limit, clipMax - 200)); + let env; + do { + const slice = full.slice(start, start + lim); + const nextOffset = start + slice.length < totalChars ? start + slice.length : null; + env = { method, title: data.title || '', offset: start, content: slice, returned_chars: slice.length, + content_chars: totalChars, total_chars: totalChars, + source_bytes: (data.source_bytes === undefined ? null : data.source_bytes), + truncated: nextOffset !== null, next_offset: nextOffset, + word_count: data.word_count || 0, method_proxy: data.method || 'extract' }; + if (JSON.stringify(env).length <= clipMax || lim <= 500) break; + lim -= Math.max(1, JSON.stringify(env).length - clipMax); + } while (true); + return JSON.stringify(env); } catch (e) { return JSON.stringify({ error: 'Fetch failed: ' + e.message }); } } case 'wpm_install': { diff --git a/modules/core/module-billing.js b/modules/core/module-billing.js index 87f8dc3..4ccf196 100644 --- a/modules/core/module-billing.js +++ b/modules/core/module-billing.js @@ -1,8 +1,10 @@ // Billing / pricing surface for newOS4 (2026-10-28) // // Why: the terminal could show a model's name but never what it costs. Prices and the anonymous -// message allowance are public data, so they need no login — the account BALANCE does, and newOS4 -// has no session layer, so that is reported honestly as "not signed in" rather than guessed. +// message allowance are public data, so they need no login — the account BALANCE does. +// 2026-11-03: newOS4 DOES have a session — the browser sends the HttpOnly cookie and +// parse_bearer_token() reads it — so the balance below is real when logged in. The old comment +// claimed otherwise and made the 401 look expected. // // Commands: cost [on|off], usage class BillingModule { @@ -56,7 +58,9 @@ class BillingModule { } async _loadBalance() { - // Only answers when a session exists; newOS4 currently has none, so this is expected to 401. + // Cookie auth is what makes this work: the browser sends webx11_session_token and + // parse_bearer_token() falls back to it. If this 401s, the user is logged out — say so, + // do not assume newOS4 cannot authenticate (that comment was wrong until 2026-11-03). try { const r = await fetch('/api/billing/subscription', { credentials: 'include' }); if (!r.ok) return null; @@ -67,6 +71,16 @@ class BillingModule { } } + async _loadUsage(days) { + try { + const r = await fetch('/api/billing/usage?days=' + (days || 1), { credentials: 'include' }); + if (!r.ok) return null; + return await r.json(); + } catch (e) { + return null; + } + } + // --- session spend, measured where the tokens actually are ---------------- _hookUsage() { const om = this.terminal && this.terminal.ollamaManager; @@ -109,8 +123,10 @@ class BillingModule { } } + // On by default since 2026-11-03: "the terminal shows tokens but not cost" was the default, + // not a bug — the flag was opt-in and nobody knew the command existed. `cost off` opts out. _costLineOn() { - try { return localStorage.getItem('webx11_cost_line') === '1'; } catch (e) { return false; } + try { return localStorage.getItem('webx11_cost_line') !== '0'; } catch (e) { return true; } } // --- the command ---------------------------------------------------------- @@ -138,13 +154,20 @@ class BillingModule { p('📊 Free allowance: unavailable (billing API did not answer)', 'warning'); } - p(`💰 Balance: ${bal && bal.balance_usd != null ? '$' + Number(bal.balance_usd).toFixed(2) : 'not signed in — newOS4 sends no session token, so a balance cannot be shown here yet'}`, 'info'); + p(`💰 Balance: ${bal && bal.balance_usd != null ? '$' + Number(bal.balance_usd).toFixed(2) : 'not signed in — log in at /spaces/dashboard.html to see your balance'}`, 'info'); if (bal && bal.plan_allowance) { const pa = bal.plan_allowance; p(`🎁 Included usage (${pa.plan}): $${Number(pa.used_usd).toFixed(2)} of $${Number(pa.limit_usd).toFixed(2)} ${pa.period === 'day' ? 'today' : 'this month'} — $${Number(pa.remaining_usd).toFixed(2)} left, then your balance pays`, 'info'); } + const u = await this._loadUsage(1); + if (u && u.totals) { + p(`📅 Today (all newOS4 sessions): ${u.totals.requests} call(s), ` + + `${((u.totals.prompt_tokens + u.totals.completion_tokens) / 1000).toFixed(1)}k tokens, ` + + `≈ $${Number(u.totals.retail_usd).toFixed(2)} retail`, 'info'); + } + if (s && s.msgs) { p(`🧮 This session: ${s.msgs} call(s), ${(s.prompt / 1000).toFixed(1)}k in / ${(s.completion / 1000).toFixed(1)}k out ≈ $${s.cost.toFixed(4)} retail`, 'info'); Object.entries(s.byModel) diff --git a/modules/core/module-fetcher.js b/modules/core/module-fetcher.js index 45505a7..d1b7bcc 100644 --- a/modules/core/module-fetcher.js +++ b/modules/core/module-fetcher.js @@ -5,7 +5,7 @@ class WebFetcherModule { } getHelpInfo() { return { - 'fetch': 'fetch <url> [--post=BODY] [--type=CT] [--raw] — Fetch URL (raw HTML/text); --post enables POST with body, --type sets Content-Type, --raw shows untruncated content', + 'fetch': 'fetch <url> [--post=BODY] [--type=CT] [--raw] [--save=FILE] — Fetch URL (raw HTML/text); --post enables POST with body, --type sets Content-Type, --raw shows untruncated content, --save writes it to the VFS', 'extract': 'extract <url> — Extract readable article content (title + body text) from a URL', }; } @@ -69,6 +69,21 @@ class WebFetcherModule { return; } const text = await resp.text(); + // --save: the only terminal path from a remote URL to the VFS. Without it a + // 100 KB source page can only be read through 3 000-char scrollback, which is + // what sent the in-OS agent looking for a shell redirect (reported 2026-11-06). + if (flags.save) { + const path = String(flags.save === true ? (url.split('/').pop().split('?')[0] || 'fetch.txt') : flags.save); + const dir = path.includes('/') ? path.slice(0, path.lastIndexOf('/')) : ''; + if (dir) { try { this.terminal.fs.mkdir(dir); } catch (e) {} } + const wrote = this.terminal.fs.writeFile(path, text); + if (!wrote || this.terminal.fs.readFile(path) !== text) { + this.terminal.print(`Save failed: could not write ${path}`, 'error'); + return; + } + this.terminal.print(`💾 Saved ${text.length} chars to ${path}`, 'success'); + return; + } const limit = showRaw ? text.length : 3000; this.terminal.print(`\n=== Content from ${url} (${method}) ===\n`, 'success'); this.terminal.print(text.substring(0, limit) + (text.length > limit && !showRaw ? '...' : '')); diff --git a/modules/core/module-loader.js b/modules/core/module-loader.js index c9162d0..1f931fd 100644 --- a/modules/core/module-loader.js +++ b/modules/core/module-loader.js @@ -4,7 +4,7 @@ // Core module bundle loaded synchronously in main.html. // This loader only handles version cache-busting, splash, terminal init, and optional ext modules. - const LATEST_VER = 'v4.6.146'; + const LATEST_VER = 'v4.6.151'; const CACHE = '?v=' + LATEST_VER; // Exposed so ext modules can register their own URL with the same cache-busting version // instead of an unversioned path the browser may already have cached (shell.js did). diff --git a/modules/core/module-orchestrator.js b/modules/core/module-orchestrator.js index 6b8ee18..05d7392 100644 --- a/modules/core/module-orchestrator.js +++ b/modules/core/module-orchestrator.js @@ -1084,6 +1084,18 @@ class OrchestratorModule { if (!task) return this.terminal.print('Usage: orchestrate <task description>', 'error'); this._delegations = (this._delegations || 0) + 1; + // Parallel fan-out bookkeeping. A sub-task carries its own id, and the batch that + // spawned it hands in a sink, so the harvest below matches results by identity — + // never by a task string that a later delegation on the same agent can overwrite. + const taskId = opts.taskId || null; + const resultSink = opts.resultSink || null; + const recordResult = (status, payload) => { + if (!taskId || !resultSink || typeof resultSink.set !== 'function') return; + resultSink.set(taskId, Object.assign({ + taskId, task, status, at: new Date().toISOString() + }, payload || {})); + }; + // Parallel fan-out: decompose and run sub-tasks concurrently if (opts.parallel) { const parts = this._decomposeGoal(task); @@ -1104,38 +1116,45 @@ class OrchestratorModule { } } - const results = await Promise.allSettled(parts.map(sub => this.delegateTask(sub.goal))); - const successes = results.filter(r => r.status === 'fulfilled').length; - const failures = results.filter(r => r.status === 'rejected').length; + const runId = 'pb' + Date.now().toString(36) + '-' + (this._parallelSeq = (this._parallelSeq || 0) + 1); + const resultSink = new Map(); + const results = await Promise.allSettled(parts.map((sub, i) => + this.delegateTask(sub.goal, { taskId: runId + ':' + i, resultSink }))); - // Result synthesis: collect outbox results from each sub-task's agent + // Result synthesis: one row per sub-task, looked up by the id handed to it. const summaries = []; for (let i = 0; i < parts.length; i++) { const sub = parts[i]; const r = results[i]; - if (r.status === 'fulfilled') { - // Find the agent that handled this sub-task - const agent = Object.values(this._agents).find(a => a.currentTask === sub.goal || (a.lastTask === sub.goal)); - if (!agent) { - // "ok / completed (agent recycled)" told the caller a task succeeded - // when we could not even find its agent. - summaries.push({ task: sub.goal, status: 'failed', error: 'agent not found — result unverifiable' }); - continue; - } - const outbox = this.terminal.fs.fsReadJSON(`/agents/${agent.id}/outbox.json`); - const body = outbox && typeof outbox.result === 'string' ? outbox.result : ''; - const snippet = body.substring(0, 200); + const subTaskId = runId + ':' + i; + const rec = resultSink.get(subTaskId); + if (rec) { summaries.push({ - task: sub.goal, agent: agent.id, - status: body ? 'ok' : 'failed', - result: body || null, - error: body ? undefined : (outbox && outbox.error) || 'no result in outbox' + task: sub.goal, + taskId: subTaskId, + agent: rec.agentId, + status: rec.status === 'ok' ? 'ok' : 'failed', + result: rec.status === 'ok' ? (rec.result || null) : null, + error: rec.status === 'ok' ? undefined : (rec.error || 'no result recorded') }); - } else { - summaries.push({ task: sub.goal, status: 'failed', error: r.reason?.message || 'unknown error' }); + continue; } + // The call died before it could report itself (a throw from a path that never + // reached a recordResult call): the promise state is the only evidence left. + summaries.push({ + task: sub.goal, + taskId: subTaskId, + status: 'failed', + error: (r.status === 'rejected' && ((r.reason && r.reason.message) || String(r.reason))) || 'no result recorded' + }); } + // The tally IS the summaries — never the promise states. allSettled counts + // resolutions, and a spawn refusal returns terminal.print() (undefined), which + // FULFILS: that is how "3/3 succeeded" printed over three failed rows. + const successes = summaries.filter(s => s.status === 'ok').length; + const failures = summaries.length - successes; + // Store merged result to VFS for persistence const merged = { type: 'parallel_result', @@ -1157,7 +1176,8 @@ class OrchestratorModule { const snippet = s.result ? s.result.replace(/\n/g, ' ').substring(0, 60) : ''; this.terminal.print(' ' + icon + ' ' + s.task.substring(0, 35) + agentTag + ' → ' + snippet, 'info'); } - this.terminal.print('✅ Parallel batch: ' + successes + '/' + parts.length + ' succeeded' + (failures > 0 ? ', ' + failures + ' failed' : ''), 'header'); + // The header must agree with the rows above it: ✅ only when nothing failed. + this.terminal.print((failures === 0 ? '✅' : '❌') + ' Parallel batch: ' + successes + '/' + parts.length + ' succeeded' + (failures > 0 ? ', ' + failures + ' failed' : ''), failures === 0 ? 'header' : 'error'); return merged; } // Single sub-task — fall through to normal path @@ -1176,8 +1196,10 @@ class OrchestratorModule { // Auto-spawn with role-derived capabilities if (!agent) { if (Object.keys(this._agents).length >= this._SPAWN_LIMIT) { - // Try finding any non-failed agent to reuse - const reuse = Object.values(this._agents).find(a => a.status !== 'failed' && a.status !== 'sleeping'); + // Reuse an agent that is NOT running. A BUSY agent must never be handed a + // second task: concurrent sub-tasks then overwrite currentTask/leaseId and + // erase each other's results (reproduced 4/4 — Defect 2b). + const reuse = Object.values(this._agents).find(a => a.status !== 'busy' && a.status !== 'failed'); if (reuse) { reuse.status = 'busy'; agent = reuse; @@ -1187,11 +1209,18 @@ class OrchestratorModule { const role = this._deriveRole(task); const caps = this._deriveCapabilities(task); agent = this.spawnAgent(role, caps); - if (!agent) return this.terminal.print('❌ No available agents — spawn limit reached', 'error'); + if (!agent) { + const msg = 'no idle agent — ' + Object.keys(this._agents).length + '/' + this._SPAWN_LIMIT + ' agents busy, refusing to share one'; + recordResult('failed', { error: msg }); + return this.terminal.print('❌ ' + msg, 'error'); + } agent.status = 'busy'; // claim immediately after spawn } } - if (!agent) return this.terminal.print('❌ No available agents', 'error'); + if (!agent) { + recordResult('failed', { error: 'no agent available' }); + return this.terminal.print('❌ No available agents', 'error'); + } // A2A Contract handshake const registry = this.terminal.moduleManager.modules['AgentRegistryModule']; @@ -1261,6 +1290,9 @@ class OrchestratorModule { // Return result for callers that await delegateTask outboxResult = _answer; // '' when the worker returned nothing; callers see that + recordResult(_answer ? 'ok' : 'failed', _answer + ? { result: _answer, agentId: agent.id } + : { agentId: agent.id, error: 'agent produced no answer' }); // Check restart policy — if always, set back to idle for next task if (agent.restartPolicy === 'always') { @@ -1272,8 +1304,10 @@ class OrchestratorModule { } else { this.terminal.print(`ℹ️ AI module not available — task queued for ${agent.id}`, 'info'); outboxResult = 'queued — AI module not loaded, nothing executed'; + recordResult('failed', { agentId: agent.id, error: 'AI module not loaded — nothing executed' }); } } catch (e) { + recordResult('failed', { agentId: agent.id, error: e.message || String(e) }); this._handleAgentFailure(agent.id, e.message); } diff --git a/modules/ext/cryptogzip.js b/modules/ext/cryptogzip.js index 200abdd..cc275df 100644 --- a/modules/ext/cryptogzip.js +++ b/modules/ext/cryptogzip.js @@ -15,7 +15,8 @@ class CryptoGzipModule { 'decrypt': 'decrypt <path.enc> <passphrase> — Decrypt a .enc file', 'tar': 'tar <src-dir> <out.tar[.gz]> — Bundle a VFS directory into a tar archive (gzip if .gz)', 'untar': 'untar <path.tar[.gz]> [dest-dir] — Extract a tar archive into VFS (auto-detects gzip)', - 'wget': 'wget <url> [filename] — Download a URL and save to VFS' + 'wget': 'wget <url> [filename] — Download a URL (via /api/fetch) and save it to VFS', + 'curl': 'curl [-s] [-o FILE] <url> — Same as wget: proxy fetch + verified VFS write' }; } @@ -342,7 +343,10 @@ class CryptoGzipModule { } // ============================================================ - // WGET — fetch URL and save to VFS (binary-safe) + // WGET — fetch a URL through the /api/fetch proxy and save to VFS + // A direct browser fetch() only works same-origin: cross-origin URLs + // throw "Failed to fetch" (CORS), measured 2026-11-06 on the served + // v4.6.150 page. /api/fetch returns the raw source as UTF-8 text. // ============================================================ async _cmdWget(args) { @@ -350,20 +354,40 @@ class CryptoGzipModule { this.terminal.print('Usage: wget <url> [filename] — Download a URL and save to VFS', 'warning'); return; } - const url = args[0]; - const filename = args[1] || url.split('/').pop() || 'download.bin'; + const url = /^https?:\/\//i.test(args[0]) ? args[0] : 'https://' + args[0]; + const filename = args[1] || url.split('/').pop().split('?')[0] || 'download.bin'; this.terminal.print(`Downloading ${url}...`, 'info'); + let text; try { - const resp = await fetch(url); - if (!resp.ok) throw new Error(`HTTP ${resp.status}: ${resp.statusText}`); - const buf = await resp.arrayBuffer(); - const bytes = new Uint8Array(buf); - this._vfsWriteBytes(filename, bytes); - const sizeKb = (bytes.length / 1024).toFixed(1); - this.terminal.print(`✅ wget: ${filename} (${sizeKb} KB, ${bytes.length} bytes)`, 'success'); + const resp = await fetch('/api/fetch?url=' + encodeURIComponent(url)); + if (resp.status === 403) { + this.terminal.print('wget error: URL blocked — internal network', 'error'); + return; + } + if (!resp.ok) { + let detail = 'HTTP ' + resp.status; + try { const j = await resp.json(); detail = j.error || detail; } catch (e) {} + this.terminal.print('wget error: ' + detail, 'error'); + return; + } + text = await resp.text(); } catch (e) { - this.terminal.print(`wget error: ${e.message}`, 'error'); + this.terminal.print('wget error: ' + e.message, 'error'); + return; + } + // writeFile() returns false rather than throwing when the parent directory is + // missing — the old code printed ✅ on that path (measured: writeFile('nosuchdir/x.txt') === false). + const dir = filename.includes('/') ? filename.slice(0, filename.lastIndexOf('/')) : ''; + if (dir) { try { this.terminal.fs.mkdir(dir); } catch (e) {} } + if (!this.terminal.fs.writeFile(filename, text)) { + this.terminal.print(`wget error: could not write ${filename} (missing directory, or path is a directory)`, 'error'); + return; } + if (this.terminal.fs.readFile(filename) !== text) { + this.terminal.print(`wget error: write verification failed for ${filename}`, 'error'); + return; + } + this.terminal.print(`✅ wget: ${filename} (${(text.length / 1024).toFixed(1)} KB, ${text.length} chars)`, 'success'); } // ============================================================ @@ -582,6 +606,22 @@ class CryptoGzipModule { this.commands['tar'] = async (args) => self._cmdTar(args); this.commands['untar'] = async (args) => self._cmdUntar(args); this.commands['wget'] = async (args) => self._cmdWget(args); + // curl — the flags an agent actually types. No shell semantics: -s/-L/-f are + // accepted and ignored, -o/--output selects the VFS target. Delegates to the + // single verified writer above so the two cannot drift. + this.commands['curl'] = async (args) => { + const positional = []; + let out = null; + for (let i = 0; i < args.length; i++) { + const a = String(args[i]); + if (a === '-o' || a === '-O' || a === '--output') { out = args[++i]; } + else if (a.startsWith('-')) { /* accepted, ignored */ } + else positional.push(a); + } + if (!positional[0]) { this.terminal.print('Usage: curl [-s] [-o FILE] <url>', 'warning'); return; } + const target = out || (positional[0].split('/').pop().split('?')[0] || 'download.bin'); + return self._cmdWget([positional[0], target]); + }; this.commandsMeta = { gzip: { argType: 'file', description: 'Compress a VFS file to .gz' }, gunzip: { argType: 'file', description: 'Decompress a .gz file (binary-safe)' }, @@ -589,7 +629,8 @@ class CryptoGzipModule { decrypt: { argType: 'file', description: 'Decrypt a .enc file' }, tar: { argType: 'file', description: 'Bundle a VFS directory into a tar archive' }, untar: { argType: 'file', description: 'Extract a tar archive into VFS' }, - wget: { argType: 'url', description: 'Download a URL and save to VFS' } + wget: { argType: 'url', description: 'Download a URL and save to VFS' }, + curl: { argType: 'url', description: 'curl [-s] [-o FILE] <url> — same as wget' } }; } } diff --git a/modules/ext/media.js b/modules/ext/media.js index 5723e27..bdc50e2 100644 --- a/modules/ext/media.js +++ b/modules/ext/media.js @@ -1,13 +1,13 @@ // newOS4 Media Analysis Module (v4.5.79) // Commands: analyze <path> [--flags] // Uses gemini-3-flash-preview:cloud for vision (images, video frames) -// Uses deepseek-v4-flash:cloud for text analysis (audio features, summaries) +// Uses deepseek-v4.1-flash:cloud for text analysis (audio features, summaries) class MediaModule { constructor() { this.commands = {}; this.terminal = null; this._visionModel = 'gemini-3-flash-preview:cloud'; - this._textModel = 'deepseek-v4-flash:cloud'; + this._textModel = 'deepseek-v4.1-flash:cloud'; this._maxFrames = 10; // max video frames to analyze this._frameInterval = 5; // seconds between video frames } @@ -118,7 +118,7 @@ class MediaModule { } // Send features to text model for AI interpretation - this.terminal.print('\n🤖 AI interpretation (deepseek-v4-flash)...', 'info'); + this.terminal.print('\n🤖 AI interpretation (deepseek-v4.1-flash)...', 'info'); const prompt = `Analyze these audio features and describe what the audio likely sounds like. Consider the genre, mood, and characteristics: Duration: ${features.duration.toFixed(1)}s diff --git a/modules/ext/webdav.js b/modules/ext/webdav.js index a97a8bc..fa42b96 100644 --- a/modules/ext/webdav.js +++ b/modules/ext/webdav.js @@ -976,12 +976,20 @@ class WebDAVModule { }; const mount = document.getElementById('wdav-tg-mount'); // A Telegram widget cannot log anyone in inside an Android WebView or an in-app - // browser: it needs a popup that can postMessage back to the opener, and those hosts - // hand the link to the system browser, which has no opener. Measured 2026-11-02 (the - // operator's phone UA ends "...; wv)"): every attempt landed in Chrome and died there. - // In that case skip the widget entirely and start the code flow, which needs no popup. - const _tgWV = /;\s*wv\)/.test(navigator.userAgent || '') || - /\bTelegram\b/i.test(navigator.userAgent || ''); + // browser: tapping it hands the whole tab to oauth.telegram.org and must come back to + // this page with #tgAuthResult. Those hosts either hand the navigation to the system + // browser (no way back) or never return. Measured 2026-11-03, operator's phone. + // Fallback only: modules/wx11-tg-webview.js is the source of truth when the host page + // loaded it (the dashboard does). This inline copy missed "Version/4.0" WebViews — the + // operator's own phone — until 2026-11-03, and the Meta/Line classes until the review + // that followed. Probe: /tg-widget-test.html. + const _ua = navigator.userAgent || ''; + const _tgWV = (window.WX11TG && window.WX11TG.isWebView) + ? window.WX11TG.isWebView() + : (/;\s*wv\)/.test(_ua) || + (/\bVersion\/4\.0\b/.test(_ua) && /Android/i.test(_ua) && !/\bSamsungBrowser\b/.test(_ua)) || + /\bTelegram\b/i.test(_ua) || + /\bFBAN\b|\bFBAV\b|\bInstagram\b|\bLine\//i.test(_ua)); const tw = document.createElement('script'); tw.src = 'https://telegram.org/js/telegram-widget.js?22'; tw.setAttribute('data-telegram-login', 'roducbot'); diff --git a/modules/test/test-orchestrator-parallel-harvest.js b/modules/test/test-orchestrator-parallel-harvest.js new file mode 100644 index 0000000..4381699 --- /dev/null +++ b/modules/test/test-orchestrator-parallel-harvest.js @@ -0,0 +1,132 @@ +// node modules/test/test-orchestrator-parallel-harvest.js +// Defect 2b (2026-11-02/03): orchestrate-parallel reported success it could not prove. +// +// Three separate lies, all in one block: +// 1. The harvest re-found each sub-task's agent by task STRING (`currentTask === sub.goal || +// lastTask === sub.goal`). An agent that had already moved on to another sub-task no longer +// carried the old goal, so a task that had demonstrably finished was dropped as +// "agent not found — result unverifiable". +// 2. The "N/N succeeded" header counted Promise.allSettled resolutions, not outcomes — a spawn +// refusal returns terminal.print() (undefined), which FULFILS, so a batch where nothing ran +// printed green. +// 3. At pool cap, delegateTask stole a BUSY agent, so two sub-tasks overwrote +// currentTask/leaseId and erased each other's results. +// +// This file pins the contract: results are harvested by sub-task id, the tally comes from the +// summaries, and a sub-task that could not get an agent is a failure, never a green row. +const fs = require('fs'), path = require('path'); +const ROOT = path.join(__dirname, '..', 'core'); +// Point ORCH_SRC at modules/core/core-bundle.js to run these assertions against the DEPLOYED +// artifact instead of the source file — a green source test says nothing about the bundle. +const SRC_FILE = process.env.ORCH_SRC ? path.resolve(process.env.ORCH_SRC) : path.join(ROOT, 'module-orchestrator.js'); + +const store = {}; +global.localStorage = { getItem: k => (k in store ? store[k] : null), setItem: (k, v) => { store[k] = String(v); }, removeItem: k => { delete store[k]; } }; +global.window = global; +global.addEventListener = () => {}; + +const Orch = new Function(fs.readFileSync(SRC_FILE, 'utf8') + '\nreturn OrchestratorModule;')(); + +let fails = 0, passes = 0; +const ok = (n, c, d) => { if (c) { passes++; return; } fails++; console.log('FAIL ' + n + (d ? ' — ' + d : '')); }; + +// A VFS stub that behaves like the real one where it matters: paths normalise, and +// fsReadJSON returns null (never throws) for a missing/corrupt file. +function mkEnv(opts) { + const prints = []; + const files = new Map(); + const norm = p => String(p).replace(/^\/+/, ''); + const term = { + prints, + print(msg, type) { prints.push({ msg: String(msg), type: type || '' }); }, + fs: { + readFile: p => (files.has(norm(p)) ? files.get(norm(p)) : null), + writeFile: (p, c) => { files.set(norm(p), c); return true; }, + readDir: () => true, + mkdir: () => true, + rm: () => true, + fsReadJSON: p => { try { const raw = files.get(norm(p)); return raw === undefined ? null : JSON.parse(raw); } catch (e) { return null; } }, + }, + moduleManager: { modules: {}, _setInterval: () => 0, _setTimeout: () => 0 }, + ollamaManager: null, + }; + const ai = { _sessions: new Map(), chat: opts.chat, terminal: term }; + term.moduleManager.modules['WebX11AIAssistantModule'] = ai; + + const orch = Object.create(Orch.prototype); + orch.terminal = term; + orch.commands = {}; + orch._agents = {}; + orch._a2aMessages = {}; + orch._eventBus = {}; + orch._restartBackoffs = {}; + orch._delegations = 0; + orch._SPAWN_LIMIT = opts.spawnLimit || 5; + orch._RESTART_POLICY = 'none'; + return { orch, term, files, prints }; +} + +const GOAL = 'read alpha log and write beta report and check gamma status'; +const tick = (ms) => new Promise(r => setTimeout(r, ms || 5)); + +(async () => { + // ── T1: three independent sub-tasks are harvested by identity, not by task string ── + { + const { orch, files, prints } = mkEnv({ + chat: async (msgs) => { await tick(); return { content: 'ANSWER[' + msgs[0] + ']', toolTurns: 0 }; }, + }); + const merged = await orch.delegateTask(GOAL, { parallel: true }); + const decomposed = prints.some(p => /Decomposed into 3 sub-tasks/.test(p.msg)); + ok('T1a the goal decomposes into 3 sub-tasks (harness precondition)', decomposed, prints.map(p => p.msg).join(' | ')); + ok('T1b all three sub-tasks are harvested', merged.summaries.length === 3, 'summaries=' + merged.summaries.length); + ok('T1c every sub-task carries its own answer', merged.summaries.every(s => s.status === 'ok' && /^ANSWER\[\[DELEGATED\]/.test(s.result || '')), JSON.stringify(merged.summaries.map(s => [s.status, s.result]))); + ok('T1d no sub-task is discarded as unverifiable', !merged.summaries.some(s => /agent not found/.test(s.error || '')), JSON.stringify(merged.summaries.map(s => s.error))); + ok('T1e successes come from the summaries', merged.successes === 3 && merged.failures === 0, 'successes=' + merged.successes + ' failures=' + merged.failures); + const onDisk = JSON.parse(files.get('orchestrator/parallel-results.json')); + ok('T1f the persisted payload agrees with the returned tally', onDisk.successes === merged.successes && onDisk.summaries.length === 3, JSON.stringify({ onDisk: onDisk.successes, merged: merged.successes })); + ok('T1g each summary is keyed by its own sub-task id', new Set(merged.summaries.map(s => s.taskId)).size === 3, JSON.stringify(merged.summaries.map(s => s.taskId))); + } + + // ── T2: at pool cap a sub-task is REFUSED, never run on a busy agent ── + // Old behaviour: all three ran on the one agent, clobbering each other, and the header + // still read "3/3 succeeded" while two summaries said failed. + { + const { orch, prints } = mkEnv({ + spawnLimit: 1, + chat: async () => { await tick(10); return { content: 'single-agent answer', toolTurns: 0 }; }, + }); + const merged = await orch.delegateTask(GOAL, { parallel: true }); + const oks = merged.summaries.filter(s => s.status === 'ok').length; + ok('T2a only the one agent that was free produced a result', oks === 1 && merged.successes === 1, 'ok=' + oks + ' successes=' + merged.successes); + ok('T2b the refused sub-tasks are reported as failures with a reason', merged.failures === 2 && merged.summaries.filter(s => s.status === 'failed').every(s => /idle|busy|limit/i.test(s.error || '')), JSON.stringify(merged.summaries.map(s => [s.status, s.error]))); + ok('T2c the tally never exceeds the green rows', !merged.summaries.some(s => s.status !== 'ok') || merged.successes < merged.total, 'successes=' + merged.successes + '/' + merged.total); + const header = prints.map(p => p.msg).find(m => /Parallel batch:/.test(m)) || ''; + ok('T2d the header reports the real outcome', /1\/3 succeeded, 2 failed/.test(header), header); + ok('T2e a batch with failures is not printed as a green header', !/^✅/.test(header), header); + } + + // ── T3: a worker that answers nothing is a failure, not a success ── + { + const { orch } = mkEnv({ + chat: async (msgs) => { await tick(); return { content: /beta/.test(msgs[0]) ? '' : 'ok answer', toolTurns: 0 }; }, + }); + const merged = await orch.delegateTask(GOAL, { parallel: true }); + const silent = merged.summaries.filter(s => /beta/.test(s.task)); + ok('T3a the silent worker is a failed row', silent.length === 1 && silent[0].status === 'failed', JSON.stringify(silent)); + ok('T3b successes count the real answers only', merged.successes === 2 && merged.failures === 1, 'successes=' + merged.successes + ' failures=' + merged.failures); + } + + // ── T4: a thrown sub-task is not swallowed, and does not inflate the tally ── + { + const { orch } = mkEnv({ + chat: async (msgs) => { await tick(); if (/gamma/.test(msgs[0])) throw new Error('worker exploded'); return { content: 'fine', toolTurns: 0 }; }, + }); + const merged = await orch.delegateTask(GOAL, { parallel: true }); + const bad = merged.summaries.filter(s => /gamma/.test(s.task)); + ok('T4a the throwing sub-task is a failed row with its error', bad.length === 1 && bad[0].status === 'failed' && /worker exploded/.test(bad[0].error || ''), JSON.stringify(bad)); + ok('T4b the tally reflects 2 real answers out of 3', merged.successes === 2 && merged.failures === 1, 'successes=' + merged.successes + ' failures=' + merged.failures); + } + + console.log('test-orchestrator-parallel-harvest: ' + passes + ' passed, ' + fails + ' failed'); + process.exit(fails ? 1 : 0); +})(); |