YouTube Comment Archive & Delta Feed

Track YouTube comment additions, edits, and removals across repeated runs with stable comment identities. The result is an append-only change feed rather than another one-time comment export.

Start in 30 seconds

1. Select Try for free and run the seeded multi-cycle demo.
2. Switch to bounded live mode or supply your own comment observations.
3. Reuse the named state store to compare later runs.

Price: $0.0005 per useful live added, modified, or explicitly removed comment, plus a $0.00005 start event. Unchanged, unavailable, failed, and demo results are free.

Seeded mode is synthetic. Live HTML collection is experimental and incomplete results never prove that a comment was deleted.

Highlights

• Append-Only Stable Identity: Every comment receives a canonical composite key youtube/<videoId>/<commentId> that prevents duplicates and tracks historical appearance across cycles.
• Deterministic Delta Engine: Classifies every comment in relation to prior state:
  • added: Newly posted comment.
  • modified: Text edited or like/reply metrics updated.
  • unchanged: Verified active with identical text and metrics.
  • removed: Comment deleted or removed since prior successful check.
  • unavailable: External access to video failed; state is held without false modifications.
• Honest Failure Semantics: Network failures, private/removed videos, or disabled comments report explicit unavailable or unknown statuses—never false "unchanged" and never false "removed".
• Credential-Free Seeded Demo: Runs out of the box with zero credentials, zero network access, and writes a non-empty dataset in seconds.

Quickstart

Run locally

To run the credential-free seeded demo:

$npm start

This completes in under 5 seconds, loads multi-cycle seeded test fixtures, executes delta evaluation, and writes the results to the Apify dataset and KeyValueStore.

Run tests

All unit, behavioral, delta precision, and error-handling tests run with Node's native test runner:

$npm test

Input Configuration

See .actor/input_schema.json for full JSON schema definitions.

Field	Type	Default	Description
targetUrls	array	["https://www.youtube.com/watch?v=dQw4w9WgXcQ"]	Target YouTube video URLs or video IDs
maxCommentsPerTarget	integer	100	Maximum comments to inspect per video
mode	string	"seeded"	"seeded" for offline deterministic demo, "live" for bounded public fetches
stateStoreName	string	mode-specific	Named KeyValueStore for persistence (null disables persistence); seeded and live defaults are isolated
emitAll	boolean	true	When true, emits all comments; when false, emits only changed deltas
priorState	object	null	Optional explicit prior state object for replay/testing

Output Datasets and Storage

• Dataset: Formatted rows following .actor/dataset_schema.json. Each row contains key, commentId, videoId, author, text, likeCount, replyCount, delta, status, targetStatus, firstSeenAt, lastSeenAt, and changes.
• KeyValueStore OUTPUT: Run envelope detailing execution timestamp, run status (ok, degraded, failed), summary delta counts, and per-target health.
• KeyValueStore STATE: Durable state record persisted in the configured store for subsequent runs.

Architecture

• src/main.mjs: Actor platform entrypoint and adapter.
• src/core/identity.mjs: Pure URL normalization and composite key generation.
• src/core/archive.mjs: Pure deterministic delta evaluator and state transitions.
• src/fetcher/youtube.mjs: Bounded read-only fetcher with explicit failure classification.
• fixtures/seeded_corpus.json: Curated multi-cycle golden fixtures for deterministic testing.

Public comments still contain author data. Set a retention period appropriate to your use, honor deletion/data-rights requests, and delete the named store and datasets when that period ends. Use one writer per target scope. No proxy, cookies, login, or access-control bypass is included. Support owner: Mehdi Badawi through the Apify Store support channel, with an initial-response target of two business days.