Add a config-driven crawl meter (1 USD / 1000 successes by default, retunable
via WEB_CRAWL_MICROS_PER_SUCCESS) mirroring EtlCreditService:
- WebCrawlCreditService (gate -> pre-check -> post-charge) on the unified wallet
- webcrawler indexer: pre-flight credit block + post-charge of the workspace
owner per success, with a web_crawl TokenUsage audit row
- chat scrape_webpage (main + research): fold each successful scrape into the
active turn's token accumulator so it settles with the chat turn
- WEB_CRAWL_CREDIT_BILLING_ENABLED / WEB_CRAWL_MICROS_PER_SUCCESS config + env
- unit tests (service) + integration tests (indexer wiring)
- scripts/e2e_phase3_crawl_billing.py: manual functional e2e (3a/3b/3c)
Co-authored-by: Cursor <cursoragent@cursor.com>
- Standardized the web crawler to use Scrapling exclusively, removing Firecrawl entirely.
- Updated the crawler's location to `app/proprietary/web_crawler/connector.py` under a non-Apache-2 license boundary.
- Refactored the `WebCrawlerConnector` to eliminate the Firecrawl API key dependency, simplifying the interface for crawling URLs.
- Adjusted related components to accommodate the new structure and ensure successful crawl outcomes are properly handled.
- Updated documentation to reflect these changes and the new implementation status.
- Updated main-agent middleware to clarify that both filesystem reads/writes and knowledge-base retrieval are handled by the `knowledge_base` subagent.
- Introduced `_forward_mention_pins` function to carry `@`-mention pins into subagent state.
- Revised system prompts to reflect the new retrieval method and ensure proper citation handling.
- Removed the `search_knowledge_base` tool and its related tests, consolidating functionality under the `task` tool.
- Enhanced documentation to guide usage of the new retrieval approach and citation practices.
The pull-based KB design (on-demand search_knowledge_base tool + pre-injected
workspace tree) fully replaced the old eager retrieval path. Remove its last
remnants:
- Delete KnowledgePriorityMiddleware (knowledge_search.py) and its tests.
- Drop the kb_priority state field + reducer default; trim
KbContextProjectionMiddleware to project only workspace_tree_text.
- Remove the now-dead feature flags enable_kb_priority_preinjection and
enable_kb_planner_runnable across backend (flags, route schema, tests,
env examples) and frontend (settings toggle, zod schema).
- Scrub <priority_documents> and stale KnowledgePriorityMiddleware references
from prompts, docstrings, and the ADR.
No functional change: nothing wrote kb_priority and neither flag gated live
behavior after the cutover. Full backend suite green (pre-existing unrelated
failures aside).
The model frequently writes citations glued to the preceding word
(docs[17]); the (?<!\w) lookbehind (added to dodge arr[1] array indexing)
silently skipped these, leaving raw [n] that fails to render and reads
like array access. Drop the lookbehind so glued citations resolve; genuine
code/array syntax stays protected by the existing code-region carve-out and
unresolved ordinals still drop harmlessly.
Rewrite the main-agent citation contract to a single [n] channel and sync
the orphaned system_prompt_composer surface to match; drop stale
[citation:chunk_id] / <chunk_index> references from dynamic_context and
provider hints. Reuse the shared hybrid search in the deliverables report
(citations omitted for now) and delete the orphaned report KB helper.
Remove the dead eager KnowledgePriorityMiddleware wiring (knowledge_priority
+ stack) and its legacy browse test. Update ADR 0001 to reflect the cutover.
web_search now registers each result as a WEB_RESULT (locator {url}) and
renders a <web_results> block of <document view="excerpt"> [n] passages,
returning Command(update={messages, citation_registry}) like
search_knowledge_base. Collapse the duplicate research-subagent web_search
into the shared tool and teach the prompts to cite web hits with [n].
The main agent's search_knowledge_base tool runs the hybrid spine, renders
a <retrieved_context> of numbered [n] passages, and persists the registry.
KB subagent prompts teach citing [n] from <document view="full"> reads
(evidence.chunk_ids -> evidence.citations). Delete the now-unused
search->read highlighting hand-off: the kb_matched_chunk_ids state field,
its reducer default, the tool's _matched_chunk_ids writer, and the dead
KnowledgePriorityMiddleware writes.
Add the checkpointed CitationRegistry (load/merge helpers + state field)
and a lightweight CitationStateMiddleware so subagents can register into
the same conversation registry. Resolve [n] -> [citation:<payload>] at
stream finalize from the registry, polymorphically by source type.
Add a shared document_render package that renders sources as
<document view="excerpt|full"> blocks with server-assigned [n] passage
labels (KB locator {document_id, chunk_id}, web locator {url}). Wire the
KB read backend (kb_postgres) and read_file to the new renderer and drop
the legacy per-document XML renderer (document_xml, retrieved_context) and
the old chunk_index / matched="true" / <chunk id> read format.
Mirror search_threads visibility in the referenced-chat resolver: a
search-space owner can now @-mention legacy threads that predate creator
tracking (null created_by_id), instead of those being silently dropped.
Add the referenced_chat_context slice: models for the data shapes, a
fail-closed resolver that fetches mentioned threads and their visible
turns under the same access rules as thread search, and a transcript
renderer that emits a budgeted <referenced_chat_context> block. When a
chat exceeds the per-reference character budget, recent turns are kept
and any leftover budget is filled with the overflowing turn's tail, with
truncation markers signalling the cut.
