feat(integrations): add Glasser block and tools

[glasserai]

Summary

Adds a Glasser integration: one block and six tools, one per Glasser GTM Solution.

Glasser is a data broker. One API key covers 1,000+ paid endpoints from providers such as Apollo, People Data Labs, Hunter, Ahrefs, Semrush, DataForSEO, Serper, Exa, ScrapeCreators and RentCast. Each Solution call names what data is wanted; Glasser routes it to a provider, falls back on provider errors, and returns the provider-native result with the exact USD charge.

Tool	Endpoint	What it does
glasser_people_search	POST /v1/solutions/gtm/people_search	Find contacts by title, seniority, location or employer; enrich one person; find a work email
glasser_company_intelligence	POST /v1/solutions/gtm/company_intelligence	Profile, tech stack, traffic, competitors, funding or news from a domain
glasser_seo_research	POST /v1/solutions/gtm/seo_research	Keyword volume and ideas, SERP, domain overview, ranked keywords, backlinks, referring domains, domain rating
glasser_web_research	POST /v1/solutions/gtm/web_research	Web, news, places, scholar, shopping, image and video search; scrape; similar pages; answer with sources
glasser_social_research	POST /v1/solutions/gtm/social_research	Search posts, read a profile, feed or single post, find other profiles on Reddit, X, YouTube, TikTok, Instagram, LinkedIn
glasser_market_data	POST /v1/solutions/gtm/market_data	US property values, rent estimates, records, listings, ZIP market stats, stock quotes

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation
• Other: ___________

Implementation notes

• Request contract: each tool is an absolute external ToolConfig.request. Every call sends a fresh Idempotency-Key (via generateId()), which the API requires; one key binds to one run, so a transport retry reads the original run instead of paying twice.
• In-flight runs: the API answers with the Run itself, complete (200) or still in flight (202). postProcess polls GET /v1/runs/{id} every 2 s for up to 3 min, tolerating up to three consecutive polling errors, following the Wiza tool's pattern.
• Outputs: the Run envelope is documented in Glasser's OpenAPI document (id, run_url, status, provider, endpoint, charge_usd, failure, task_id) and is mapped field by field. output is declared as bare json on purpose: it is the provider-native payload, and its shape depends on which provider and endpoint served the call. Nothing is guessed.
• Model input: only the answer action of glasser_web_research hands the query to a model (Exa's answer endpoint), so request.modelInput projects query for that action alone, like Linkup's sourced-answer projection.
• Block: one operation dropdown keyed by tool id, per-operation action dropdowns, a provider dropdown whose options derive from the selected operation, canvas sentences for all six operations, eight templates and four skills grounded in tools.access.
• Icon: Glasser's mark from glasser.ai; the body follows currentColor, the visor keeps its brand orange, so it renders bare in light and dark mode.
• Baseline: the knowledge page graph was one module over its check-tool-registry-boundary allowance after adding the nine tool modules, so its entry alone was re-recorded. Staging already sits close to the limit on that page.
• No hosted key, BYOK or OAuth: users bring their own Glasser API key.

Testing

• bun run tool-metadata:generate, bun run scripts/generate-docs.ts, bun run generate:block-successors, bun run deployment-config:generate — generated artifacts committed
• bun run check:audits — all 47 audits pass (includes docs:check, integration-catalog:check, check:canvas-sentences, check:bare-icons, check:tool-request-boundary, check:tool-registry-boundary, check:icon-path-precision, check:byok-providers)
• bun run apps/sim/scripts/check-canvas-sentences.ts --block=glasser — 6/6 operations
• bunx vitest run tools/glasser/glasser.test.ts — 11 tests: request headers and body shaping, run mapping, failed-run reporting, polling to completion, polling give-up, block param folding
• bunx vitest run blocks tools/model-input-selectors.test.ts tools/operation-input.test.ts — 1595 tests pass
• tsc --noEmit — no errors in the added files
• Request and response shapes follow Glasser's published OpenAPI document and match Glasser's shipped n8n and Dify adapters.

Reviewers may want to focus on apps/sim/tools/glasser/run.ts (idempotency header, polling, run mapping) and the block's tools.config.params, which folds the per-operation action dropdowns into the API's action field.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

🤖 Generated with Claude Code

[vercel]

@glasserai is attempting to deploy a commit to the Sim Team on Vercel.

A member of the Team first needs to authorize it.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the remaining stale-parameter finding is fully fixed and no new actionable issues were identified.

Summary

Adds the Glasser integration as a workflow block with six GTM data tools, API-key authentication, asynchronous run polling, provider-native outputs, catalog documentation, icons, generated metadata, deployment configuration, and regression coverage.

• Registers people search, company intelligence, SEO research, web research, social research, and market data operations.
• Maps operation-specific block controls into each tool’s request contract while preventing values from previously selected operations from leaking into later requests.
• Polls in-flight Glasser runs with bounded error handling and exposes run status, provider, charge, failure, and result data.
• Updates integration discovery surfaces and generated registries for the new block and tools.
• Adds tests covering request construction, response mapping, polling behavior, and operation parameter folding.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Glasser workflow block] --> B{Selected operation}
  B --> C[Operation-specific parameter fold]
  C --> D[Registered Glasser tool]
  D --> E[POST GTM solution endpoint]
  E --> F{Run status}
  F -->|Completed or failed| G[Map run envelope and provider output]
  F -->|Queued or running| H[Poll run endpoint with bounded retries]
  H --> F
  G --> I[Workflow output]

Loading

_{Reviews (3) · Last reviewed commit: "fix(glasser): fold only the selected ope..."}

[glasserai]

Addressed all five findings in 37b8a66 (details on each thread). bun run check:audits passes again and the Glasser suite is now 12 tests.

@greptileai

[glasserai]

Fixed the remaining finding in 3bf514d (see thread); audits and the 13 Glasser tests pass.

@greptileai