firecrawl: sync submission next_steps with the shipped catalogue metadata - #96
Open
Alexgodoroja wants to merge 3 commits into
Open
firecrawl: sync submission next_steps with the shipped catalogue metadata#96Alexgodoroja wants to merge 3 commits into
Alexgodoroja wants to merge 3 commits into
Conversation
…data The graph was expanded from 10 to 18 edges after #95 merged, so the merged submission no longer matched what the catalogue actually ships. Left as-is, a re-publish from this submission would silently regress the graph and drop the 429 concurrency edge - the single most important one, since the shared Firecrawl account runs at maxConcurrency 2. Adds: 429 (with interact-session slot recovery), match-on-success-body edges for a still-running crawl and an empty search, site-refused scrape detection, and a timeout edge. Every command was executed against the live app. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The demo is printed at the last step of install and rendered on the website, so it was publishing our cost basis with Firecrawl: the micro-USD ledger unit, the '1 credit ~= 830 micro-USD' conversion, per-operation dollar prices, and the $0.0415 budget. It also leaked that Pilot runs one shared Firecrawl account. Restated in the published limits instead: 1000 credits, 2 concurrent calls. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The next-steps hint said 'the shared account runs at maxConcurrency 2', which tells every agent that all Pilot users sit behind one Firecrawl team. Restated as the user's own published limit: 2 concurrent calls.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-up to #95. Found while reviewing before sign-off.
The drift
The
next_stepsgraph was expanded from 10 to 18 edges after #95 merged, so the merged submission no longer matches what the catalogue actually ships (pilot-protocol/pilotprotocol#447).That is not cosmetic.
submission.jsonis the authoring source of truth —publish-rich-from-r2.shcarriesnext_stepsfrom the submission intometadata.json. So a re-publish from this submission, e.g. once the brokenCATALOG_PUBLISH_TOKENis rotated andpublish-on-mergestarts working again, would silently regress the graph and drop the429edge.That edge is the most important one in the graph: Pilot's shared Firecrawl account runs at
maxConcurrency: 2, and 429 is by far the most likely failure an agent meets.What this adds
Verified identical to the graph shipped in #447 (
json.dumps(..., sort_keys=True)equal).429— routes toqueue_status, theninteract_list, because browser/interact sessions each hold one of the 2 concurrency slots until destroyed. Found live, not in the docs.match-on-success-body edges — a crawl still"scraping"is normal, not a failure; an emptyweb[]on search means fix the query, not scrape the results.statusCodeis 401/403/429/5xx — Firecrawl succeeded, the site refused; retry withproxy:"enhanced".activity, so the agent checks whether the job landed before paying to run it again.Testing
TestAllSubmissionDemosValid,TestAllSubmissionNextStepsValid,TestSubmissionGatewaysAreReachable,TestSubmissionCrossAppStepsResolveall pass.Per rule 5 of
NEXT-STEPS-GRAPHS.md, every authored command was executed against the live app, and rendering was confirmed on a realpilotctl appstore installfor all three edge classes:No bundle change, no version bump —
next_stepslives in metadata, not in the bundle.🤖 Generated with Claude Code