ナビゲーションにスキップ

Why are fewer Instagram comments exported than shown, and how do I report coverage?

Report the Instagram comments you actually retrieved instead of subtracting export rows from the displayed comment total. Count three collections separately by ID: top-level comments from instagram.fetch_post_comments_v2, reply previews inside those comments, and replies fetched with instagram.fetch_comment_replies. Take the pages checked and the stop reason from coverage.json.

For example, if the post shows 100 comments and your export holds 30 distinct top-level comments plus replies to two questions, report the 30 top-level comments and the two reply branches as separate observations. Keep the displayed total of 100 as its own observation with its observation time.

What are the three comment collections?

MaterialLocation or outputCounting method
Root commentsdata.comments, exported to comments.csvDeduplicate observed root-comment IDs
Reply previewspreview_child_comments, saved separately as JSONLKeep as preview material, not a fully read thread
Retrieved repliesdata.child_comments, exported to replies.csvRecord the selected root’s branch separately

Root records include pk, text and the optional comment_like_count / child_comment_count. Record the reported reply count in child_comment_count separately from the number of replies actually saved.

How do I continue root comments?

Keep the same media and the same sort_order (recent or popular), and pass the returned data.next_min_id unchanged as min_id. data.next_min_id is a JSON-shaped string; the helper in the export guide builds the next request from the saved response, so you never rebuild the string by hand.

Check the page count, stop reason and records in coverage.json. When the response reports a final page, the root-comment branch is complete. When the run stops at a planned page cap, the cap defines the reviewed scope. When continuation data is incomplete or a request fails, keep the material already saved.

How do I add reply context to a question?

Select a root comment from the saved original response and retrieve the root comment’s reply branch. Reply paging uses data.has_more_tail_child_comments, and the input contract names data.page_info.next_min_id as the continuation cursor. The return description also mentions next_min_child_cursor. The export guide’s helper does not treat next_min_child_cursor and data.page_info.next_min_id as interchangeable, and stops a reply branch when the required cursor is missing.

After reading the saved replies, label each question: answered with the verbatim reply as evidence, needs_context when the material is insufficient, or no_answer_observed when the sample shows no answer. Deliver resolution.json with the evidence and the collection scope.

How do I describe the result?

Write, for example: “This run saved three root-comment pages and N unique IDs, plus reply observations for two selected questions. Times and stop reasons are attached.” Replace N and the page counts with the actual records. Keep the displayed total as a separate observation with its time.

Root comments, replies and details are charged at the current pricing. Deduplication, review and organization of saved material run locally.

Capability references: data:instagram.fetch_post_comments_v2, data:instagram.fetch_comment_replies.

Complete comment and reply workflow