Update deep research agent template dashboard

[shriyatheunicorn]

Summary

• update the deep research agent dashboard to match the Browserbase Vercel preview styling
• add the Browserbase logo asset, simplified topbar branding, static pipeline details, and clearer raw output labeling
• keep direct file:// opens pointed at local dev API candidates instead of an external deployment
• remove the root index.ts Vercel function config so the static public/ UI plus /api/research deploy cleanly in the Browserbase Vercel project
• address Cursor Bugbot findings around async error handling, local concurrency, static logo serving, verifier summary consistency, source-domain caps, and failed pipeline styling
• remove the old hidden sidebar markup and dead sidebar interactions

Verification

• npm run check
• node scripts/check-readme-template-index.mjs
• inline dashboard script syntax check
• local server smoke check for /, /api/health, and /browserbase-logo.svg

---

Note

Medium Risk
Large new template adds a full research pipeline plus serverless/local HTTP endpoints that call external Browserbase/Stagehand services, so failures or misconfiguration can affect runtime cost and reliability. Changes are mostly additive and isolated to the new typescript/deep-research-agent directory plus a README index entry.

Overview
Adds a new typescript/deep-research-agent template that runs an iterative search → fetch → browser fallback → synthesis → verification research loop, persisting run workspaces (plan.md, strategy.md, traces, reports) and returning a structured JSON/Markdown brief with source IDs, claim maps, and verifier output.

Includes a static dashboard UI (public/index.html + logo) with local server support (src/local-web.ts) and Vercel serverless handlers (api/health.ts, api/research.ts, vercel.json) that expose /api/research and health checks, plus a benchmark mode that loads TSV/JSON/JSONL tasks and writes aggregate results.

Updates the root README.md template table to list deep-research-agent, and adds template-level project setup (package.json, .env.example, .gitignore) and response utilities in src/api-utils.ts.

^{Reviewed by Cursor Bugbot for commit 7f339a5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[shriyatheunicorn]

bugbot run verbose=true

[cursor]

Bugbot request id: serverGenReqId_99a1e2b6-44c5-4517-8468-61436b1da0dc

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 7f339a5. Configure here.}

[cursor]

Failed pipeline step highlights wrong stage after cycling

Low Severity

The loadingStepIndex increments every 3.6 seconds and wraps via % pipelineSteps.length (6). After ~21.6 seconds the index cycles past all steps. On failure, the expression (loadingStepIndex - 1) % pipelineSteps.length wraps to a low value — e.g., when loadingStepIndex is 7, the result is 0, marking "Planning" as failed instead of the later stage where the run actually was. Since the estimated run time is 240 seconds, most real failures would highlight the wrong pipeline step. Using Math.min(pipelineSteps.length - 1, loadingStepIndex - 1) instead of the modulo would cap at the last step for long runs.

Additional Locations (1)

• typescript/deep-research-agent/public/index.html#L1424-L1431

 

^{Reviewed by Cursor Bugbot for commit 7f339a5. Configure here.}

[jay-sahnan]

This file is quite long and would be very hard for a beginner to follow or understand what's going on. Is there any way we can split this up into separate files?