Skip to main content

Dashlink API Documentation — Completeness & SDK Cross-Check Report

Date: 2026-11-19 Scope: dashlink-api-docs (Mintlify) + all six SDK codebases in Dashlink SDK/ Question asked: Is the documentation complete and developer-friendly enough that a new developer has a full reference to integrate Dashlink over REST and via the SDKs — and does the SDK code match what the docs describe?

Verdict

Yes — the documentation is complete, developer-friendly, and accurate to the code. A new developer onboarding onto Dashlink has what they need to integrate both the REST API and the native SDKs. REST coverage is exact (84/84 endpoints), the foundational concept pages are excellent, and an independent cross-check of all six SDK codebases confirms the documented public API, method signatures, link-object fields, and install coordinates match the real code. There is a short list of fixable discrepancies. Two are genuine and should be fixed; the rest are minor polish.

Part 1 — REST API & guides

Strengths

  • Complete, verifiably in-sync REST coverage. 84 endpoints with an exact 1:1 match across three sources: the nav (docs.json), the OpenAPI spec (api-reference/openapi.json), and the per-group overview tables. Zero missing, zero orphaned. This is enforced by generator scripts (scripts/generate_nav.py, scripts/generate_group_pages.py) that derive everything from one manifest, with each operation pinning its page URL via x-mint.href and scripts/check_links.py + --check drift detection guarding against rot.
  • Excellent foundational concept pages, each with working code in six languages (curl/Node/Python/Go/Ruby/PHP):
    • authentication — key types (dlk_live_ vs dlk_sdk_), scopes, per-module permissions, the Authorization: Bearer gotcha, rotation-by-overlap.
    • errors — single envelope, full code table, the critical “bulk calls return 200 even when items fail” semantics, a real retry policy.
    • rate-limits, pagination (incl. offset-consistency caveat), versioning (clear stability contract).
  • Strong quickstart (create a link → read its analytics, ~5 min).
  • Good ergonomics: interactive playground, auto-generated 6-language samples, llms.txt, changelog, contextual “open in ChatGPT/Claude/Cursor/MCP” options, and a 58-page Help Center for the dashboard side.

Gaps (REST)


Part 2 — SDK code vs. docs cross-check

Each SDK’s actual source was diffed against its guide page (guides/app-dynamic-links/*.mdx). Across all six, the documented public API names, method signatures, all 11 link-object fields, install coordinates, and native-SDK wrapping match the code. The SDK code is correctly wired and accurately described.

Per-SDK results

Discrepancies to fix

Note on S2 (Unity): this is the one place the docs describe behavior the code does not deliver. The fix is primarily in the Unity SDK (send OnDynamicLinkError on the iOS failure path, and either wire up or remove the dead LinkFailed/LinkReceived C# events), after which the doc is already correct.

  1. Fix S1 (iOS GitHub org) — one-line-per-occurrence doc edit; also correct the iOS repo’s podspec. Highest impact, lowest effort.
  2. Fix S2 (Unity iOS error path) — code fix in the Unity SDK; confirm the doc afterward.
  3. Fill the REST field-level reference (R1) — add descriptions + examples to the OpenAPI spec’s request fields and query params. This is the single change that most raises day-to-day developer-friendliness.
  4. Add S3/S4/S5/S6 polish — Unity install specifics, soften iOS “rejects” wording, add “handled by the underlying SDK” notes to wrapper pages, document sdkVersion and the full link-field set.

Later plan (separate effort)

  • Publish and list each SDK on its respective package manager. Alongside correcting the repo references (S1), the SDKs should be published to — and documented with — their standard registries, so developers install via the idiomatic tool for each platform rather than a raw git URL: Tracked as a separate follow-up plan — the install sections of each SDK guide update once the packages are actually published.

How this was assessed

  • REST: endpoint/nav/spec cross-count, per-operation schema/description/example coverage, x-mint.href link-integrity verification, and a read of the quickstart + all five concept pages + representative guides.
  • SDKs: six parallel code reviews, one per SDK, each diffing the public API surface, install coordinates, version, and link-object fields against the corresponding guide page with file:line evidence.
All review was read-only; no code or docs were modified.