Repository navigation
Conversation
…#414) The first T7 slice. The engine has no research job, source store or citation producer, so this shows this session's WebSearch and WebFetch calls from the transcript and claims nothing beyond them. - application::research pairs each search or fetch tool call with its result by tool-use id, indexing results once so long transcripts stay linear. A call is returned, failed (one row, with the HTTP status a fetch error reports), pending, or interrupted (from the loop's interrupted tools). - Search hits are parsed only from the numbered "N. **title**" line directly followed by "URL:"; any other URL line, including one inside a snippet, is not a hit. Hits are labeled "parsed, unverified", and duplicates are kept. - URLs are displayed without credentials, query or fragment, which can carry secrets. All source text is terminal-safe, kept to one row, and never interpreted. - The inspector gains a Research page after Work. Its header says the rows are tool activity, not a research job, and that jobs, citations and export need a research producer. Each row shows retrieval time, final URL and content hash as unknown, and a returned fetch as possibly cached or model-extracted, pointing at #433. The Views entry for Research reads "local only". Tests (seen failing first against a stub): projection unit tests for pairing and order, hit parsing, failure, pending and interrupted states, URL redaction, hostile text and fake URL lines; Next render tests for content, redaction, scope notes under scrolling, no "verified" claim, empty and interrupted states, the Views entry and 40x12, 20x6 and 1x1 rendering.
…nding (#414) - Failed fetch errors carry the raw URL; every scheme:// token is now shown in display form, redacted before the text is shortened. - A backslash ends the host, as it does for the HTTP client. - Credentials and query are stripped for every scheme, not only http(s). - A call with no result and a later message reads "no result recorded". - A returned search with no parsed hits says the format was not recognised. - Only assistant tool uses are listed; fetch bodies are no longer copied. - Scope notes fit 40 columns so rows stay visible at 40x12.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
4 of 10 tasks
6 tasks done
This branch was successfully deployed
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.
Context
WebSearchandWebFetchcalls from the transcript. It says plainly that they are tool activity, not a research job, and makes no claim that anything was verified.mainas the stack lands.Implementation
crates/tui/src/application/research.rs): pairs each assistantWebSearch/WebFetchcall with its result by tool-use id, in call order. It takes one linear pass over the results.scheme://token redacted the same way, and they are redacted before they are shortened.N. **title**/URL:form, and every hit is labelled "parsed, unverified". A returned search with no parsed hits, such as DuckDuckGo's format or "No results found.", says "none parsed (output format not recognised)".Verification
cargo fmt --all --checkcargo clippy --workspace --all-targets -- -D warningscargo test --workspace: 2898 passed, 0 failed, 52 ignoredpython3 scripts/check-secrets.py: this repository has no such script. The diff adds no credentials, tokens or private paths.next_shell: 18 tests, still passingIndependent review
A fresh reviewer read
c9dc1c2. Verdict: ready after fixes. Resolved inef93e74:Risk and Rollback