Commit graph

3222 commits

Author SHA1 Message Date
Thibault Jaigu
2ff7ea4cb6 feat: add Requesty as a model provider
Add Requesty (https://requesty.ai), an OpenAI-compatible LLM router, as a
model provider by mirroring the existing OpenRouter integration.

Backend:
- app/services/requesty_model_normalizer.py: normalizes Requesty's /v1/models
  catalogue, mapping its flat capability booleans (supports_tool_calling/
  supports_vision/supports_image_generation) and context_window field onto the
  shared normalized shape (Requesty differs from OpenRouter's architecture +
  supported_parameters + context_length layout)
- provider_registry.py: Requesty ProviderSpec (OpenAI-compatible, base URL
  https://router.requesty.ai/v1, REQUESTY_API_KEY bearer auth)
- model_connection_service.py: key verification + live model discovery
- quality_score.py: Requesty score entry
- unit tests mirroring the OpenRouter normalizer coverage

Frontend:
- Requesty provider icon + registration, metadata entry, and base-url hint

Signed-off-by: Thibault Jaigu <thibault.jaigu@gmail.com>
2026-07-13 09:42:30 +01:00
Anish Sarkar
e38ca19b18 Merge remote-tracking branch 'upstream/dev' into feat/instagram-scraper 2026-07-11 04:31:11 +05:30
Anish Sarkar
819486ac46 docs(instagram): document PolarisMedia extraction and richer media fields 2026-07-11 03:44:38 +05:30
Anish Sarkar
75c3addd2b test(instagram): cover PolarisMedia relay parsing and enriched media fields 2026-07-11 03:44:33 +05:30
Anish Sarkar
d3c65a37b1 refactor(instagram): drop unused slug field and mark share links unsupported 2026-07-11 03:44:28 +05:30
Anish Sarkar
4bc3d43b56 refactor(instagram): remove unused share-link redirect resolver 2026-07-11 03:44:23 +05:30
Anish Sarkar
452fe30aa4 refactor(instagram): drop login-walled comment fields and note search_type aliasing 2026-07-11 03:44:15 +05:30
Anish Sarkar
cb4adab852 feat(instagram): extract PolarisMedia relay JSON with carousel, tags, and location 2026-07-11 03:44:10 +05:30
Anish Sarkar
69bf748c2c chore(.gitignore): ignore Instagram tagged test fixture 2026-07-11 03:25:39 +05:30
Anish Sarkar
b1fe739308 docs(instagram): update platform scraper README 2026-07-11 03:25:36 +05:30
Anish Sarkar
c001f4b16e feat(subagent): update Instagram subagent system prompt 2026-07-11 03:25:25 +05:30
Anish Sarkar
27d22a9a2a refactor(instagram): drop mentions from scrape capability 2026-07-11 03:25:20 +05:30
Anish Sarkar
5cac6612c3 refactor(instagram): update platform schemas and scraper for tagged media 2026-07-11 03:25:17 +05:30
Anish Sarkar
457be1871c feat(instagram): improve parsing of Instagram media IDs and mentions
- Enhanced the regular expression for mentions to prevent trailing punctuation from being included in handles.
- Added support for extracting media IDs from deep-link meta tags in anonymous posts.
- Updated unit tests to validate the new media ID extraction and ensure proper handling of mentions.
2026-07-11 03:06:04 +05:30
Anish Sarkar
d41ccb7edd feat(instagram): enhance Open Graph parsing for anonymous posts
- Added support for extracting likes, comments, username, timestamp, and caption from Open Graph meta tags.
- Implemented fallback mechanisms to ensure graceful degradation when expected data is missing.
- Updated unit tests to validate new parsing logic and ensure robustness against unrecognized formats.
2026-07-11 02:56:29 +05:30
Anish Sarkar
467f51fd5c chore(.gitignore): add Instagram test fixtures for unit tests 2026-07-11 02:50:43 +05:30
Anish Sarkar
4bd3ec275c delete(instagram): remove obsolete post fixture for Instagram tests 2026-07-11 02:50:26 +05:30
Anish Sarkar
949cc1b29f chore(scripts): update Instagram e2e scraper script 2026-07-11 02:21:26 +05:30
Anish Sarkar
ba70ae1f45 test(instagram): update platform tests and add discovery fixtures 2026-07-11 02:21:14 +05:30
Anish Sarkar
82a63043f7 test(instagram): update capability tests for schema changes 2026-07-11 02:21:11 +05:30
Anish Sarkar
9e9dd8a124 feat(subagent): update Instagram subagent prompts and tools 2026-07-11 02:21:02 +05:30
Anish Sarkar
36bb4a1bea docs(instagram): update platform scraper README 2026-07-11 02:20:54 +05:30
Anish Sarkar
4813dc96e3 refactor(instagram): streamline scraper, fetch, and parsing pipeline 2026-07-11 02:20:51 +05:30
Anish Sarkar
de1990f9f6 refactor(instagram): simplify scrape and details capability schemas 2026-07-11 02:20:46 +05:30
Anish Sarkar
8afa4c6fc6 refactor(instagram): remove unused comments capability 2026-07-11 02:20:39 +05:30
CREDO23
1911b78379 fix(tiktok-agent): drop keyword search from the routing description's video sources 2026-07-10 17:40:04 +02:00
CREDO23
cd82f861de fix(tiktok-agent): stop steering the subagent to keyword search for videos
The playbook listed search_queries as a video source in three places, then contradicted itself. Replace with action-oriented guidance: videos via hashtags/URLs, accounts via user_search. Drop 'login-walled' (the agent has no login path).
2026-07-10 17:40:04 +02:00
CREDO23
9793751b44 docs(tiktok): trim stage 2 comment to the why 2026-07-10 17:28:47 +02:00
CREDO23
f74a73efdb docs(tiktok): trim fetch_item_list retry docstring to intent 2026-07-10 17:28:47 +02:00
CREDO23
01bc3f8de3 docs(config): trim narration from xvfb + listing-retry comments 2026-07-10 17:28:47 +02:00
CREDO23
97e250be4b docs(env): document TIKTOK_LISTING_MAX_ATTEMPTS 2026-07-10 17:25:48 +02:00
CREDO23
8f825b7b3c test(tiktok): cover item_list retry-on-empty loop 2026-07-10 17:25:48 +02:00
CREDO23
c8fcf87685 fix(tiktok): retry empty item_list captures on a fresh exit IP
The profile feed is withheld from flagged IPs and the proxy rotates per request, so re-fetching an empty capture (up to TIKTOK_LISTING_MAX_ATTEMPTS) turns a bad first draw into a hit instead of an ErrorItem.
2026-07-10 17:25:48 +02:00
CREDO23
15fb50c83e feat(tiktok): add TIKTOK_LISTING_MAX_ATTEMPTS knob
Bounds the browser-listing retry-on-empty; default 3, set 1 for a static IP.
2026-07-10 17:25:48 +02:00
CREDO23
493579913e docs(tiktok): note stage 2 profile feed is headful only when flag is set 2026-07-10 17:17:53 +02:00
CREDO23
c29b7f4b6b docs(env): document CRAWL_HEADED_XVFB_ENABLED 2026-07-10 17:17:53 +02:00
CREDO23
d3a6e2cd45 docs(docker): note entrypoint wires xvfb at runtime
The xvfb runtime path is now implemented, so drop the stale 'deferred to Slice B' note.
2026-07-10 17:17:53 +02:00
CREDO23
6400dc5f04 fix(tiktok): gate headful profile feed on CRAWL_HEADED_XVFB_ENABLED
fetch_item_list goes headful only when the flag promises a display; otherwise it stays headless so the browser launch never fails and the empty feed degrades to an ErrorItem. Verified under xvfb-run: real videos returned on a clean proxy IP, graceful degrade on a flagged one.
2026-07-10 17:17:45 +02:00
CREDO23
a551dd7553 feat(docker): start Xvfb when CRAWL_HEADED_XVFB_ENABLED is set
For every browser-running role (api/worker/all, not migrate) the entrypoint brings up a virtual display and exports DISPLAY, registering Xvfb for graceful shutdown. Off keeps browsers headless.
2026-07-10 17:17:36 +02:00
CREDO23
6784503414 feat(crawl): add CRAWL_HEADED_XVFB_ENABLED flag
Single switch promising an Xvfb display so the stealth browser can run headful (TikTok's profile feed is empty to headless Chromium). Defaults FALSE.
2026-07-10 17:17:27 +02:00
CREDO23
44966315f1 docs(tiktok): update e2e stage 2 notes for headful profile feed
Profile videos now return via the headful item_list capture; reword the
stage 2 comments to expect real videos, ErrorItem only as fallback.
2026-07-10 16:45:53 +02:00
CREDO23
c679f2a3ef fix(tiktok): fetch profile video lists headful + dismiss login modal
The profile feed (/api/post/item_list) returns an empty 200 to headless
sessions but serves data headful on the same proxy IP. Run fetch_item_list
headful and dismiss the login modal that blocks mid-scroll.
2026-07-10 16:45:45 +02:00
CREDO23
867e940911 Merge upstream/dev into feature/tiktok-scraper
Sync with the advanced dev tip so the PR compares against current dev
(drops already-merged commits from the diff). No history rewritten.
2026-07-09 20:15:20 +02:00
CREDO23
bcfc3a5c3a test(capabilities): expect export_run in the agent tool manifest
The run-reader set gained export_run (CSV export of a stored run), but this
manifest assertion still expected only read_run/search_run. Add export_run so
the agent-door test reflects the shipped readers.
2026-07-09 20:04:30 +02:00
CREDO23
6b15d610d9 feat(tiktok): surface comments, user_search, trending across all product surfaces
Brings the three newer verbs to parity with tiktok.scrape everywhere it was
wired: MCP tools (+ selfcheck manifest), the chat subagent prompt/description,
the playground catalog, the native docs page, and the SEO marketing page. Also
adds live e2e stages for comments, user search, and trending. Docs/FAQ now
state the real contract: profile metadata is reliable while its video list can
be withheld, and keyword video search is walled (use user search for accounts).
2026-07-09 19:09:00 +02:00
CREDO23
67b5472b9f feat(tiktok): add tiktok.trending verb for the Explore feed
The Explore feed (/api/explore/item_list) is a global trending-video feed
served to anonymous sessions, and it returns the same itemStruct shape as the
other listings — so the verb reuses parse_video, the listing flow, the
TikTokVideoItem output, and the per-video billing meter wholesale. Adds a
browser-capture marker + fetch_trending, a synthetic-target orchestrator entry,
and the tiktok.trending capability, surfaced on the chat subagent.
2026-07-09 18:52:56 +02:00
CREDO23
7723b5b8b6 feat(tiktok): add tiktok.comments verb for public comment scraping
Comments load over a signed /api/comment/list XHR that TikTok serves to
anonymous sessions once the comments panel is opened (unlike profile-video and
general-search feeds), so this is a reliable verb. Given video URLs it returns
CommentItems (text, author, likes, reply counts; replies carry repliesToId),
deduped per video, capped, and degraded to an ErrorItem for empty/withheld
videos or a bad_url ErrorItem for non-video inputs.

Generalizes the browser capture over a pluggable interaction step so the
comments flow (open panel, scroll the panel to paginate) reuses the same
warm+capture scaffolding as listing/user-search. Billed per comment on a new
TIKTOK_COMMENT meter (TIKTOK_MICROS_PER_COMMENT, matching the per-comment
market), surfaced on the chat subagent alongside tiktok.scrape/user_search.
2026-07-09 18:35:26 +02:00
CREDO23
192b6dc31a feat(tiktok): add tiktok.user_search verb for account discovery
Video/general search is login-walled for anonymous sessions, but the Users
tab (/api/search/user) returns public account records without a redirect, so
this exposes the one reliably-unblocked search path. A keyword yields
TikTokProfileItems (name, followers, bio, verification), deduped per query,
capped, and degraded to an ErrorItem when a query is empty/withheld.

Reuses the browser capture (generalized over XHR markers + extractor) and the
shared profile item shape. Billed per account on a new TIKTOK_USER meter
(TIKTOK_MICROS_PER_USER), surfaced on the chat subagent alongside tiktok.scrape.
2026-07-09 18:00:40 +02:00
CREDO23
6652efd035 feat(tiktok): return profile metadata even when video listing is gated
A profile's account data (name, followers, bio, verification) lives in the
page's rehydration blob and loads over plain HTTP without a signed request,
so emit it first and always. The video listing needs a signed item_list XHR
that TikTok withholds from anonymous sessions, so it stays best-effort and
degrades to an ErrorItem. A blocked profile now yields its metadata instead
of only an ErrorItem.
2026-07-09 17:48:19 +02:00
Anish Sarkar
2c251f9e07 test(instagram): add tests for anonymous profile and hashtag discovery
Introduced new tests to validate the behavior of the Instagram scraper when discovering anonymous profiles and handling hashtag searches. The tests ensure that profile lookups succeed without requiring authentication, while hashtag searches correctly raise an error when access is blocked, enhancing the robustness of the scraper's functionality.
2026-07-09 20:26:18 +05:30