The acquisition system and the connector standard.
The acquisition system and the connector standard: the 28 calls this system serves, each a route of the API and a tool of the MCP server, at contract 0476e35e6e275db5.
Figure 1The acquisition system and the connector standardspecified
The acquisition system and the connector standard
The 28 calls below are the whole of this system on the wire: each is one route of the API and one tool of the MCP server, with one typed input, one card as its answer and one refusal shape. Every answer carries its foot: what was read, the population and the control.
The calls#
| tool | route | takes | gives |
|---|---|---|---|
| si.ask.get_content | POST /v1/ask/get_content | a ContentRequest |
a ContentResponse, the page in bronze |
| si.ask.propose_template | POST /v1/ask/propose_template | a ProposeTemplateInput |
a draft template with fields and evidence spans for the person to confirm |
| si.ask.read_source | POST /v1/ask/read_source | the installed connector, the kind, a cursor | a page of source objects, the next cursor, the cells credited |
| si.ask.scrape | POST /v1/ask/scrape | a ScrapeRequest |
a ScrapeResult, the rows as source objects with their evidence spans |
| si.ask.search | POST /v1/ask/search | a SearchRequest |
SearchResult rows, streamed as engines answer, credited to cells |
| si.read.browse_view | POST /v1/read/browse_view | a BrowseViewInput |
one frame of the page as the person sees it, with its size and address, for a dashboard that shows the session live and forwards the person’s clicks and words as steps |
| si.read.catalogue | POST /v1/read/catalogue | a tenant or a brand | every source and connector with its kinds of reading and action, its state, the identity it needs, its cost, its rights, its coverage and its health |
| si.read.coverage_by_cell | POST /v1/read/coverage_by_cell | a source, or a scope with a window | measured, never measured with the reason, withheld |
| si.read.lineage | POST /v1/read/lineage | an artefact | its whole history and footprint, marking unavailable what a store cannot answer |
| si.read.page_changes | POST /v1/read/page_changes | a PageChangesInput |
PageChange rows for the site or the watch in the window |
| si.read.scrape_runs | POST /v1/read/scrape_runs | a ScrapeRunsInput |
ScrapeRun rows with pages, rows, bytes, cost and time |
| si.read.source_procedure | POST /v1/read/source_procedure | a source | its procedure at its version, the date of its last reading and whether it is stale |
| si.act.act_through_connector | POST /v1/act/act_through_connector | the installed connector, the kind, the arguments with their sources | a move, proposed or running by the grant’s mode; then the source’s own id for what was done |
| si.act.browse_step | POST /v1/act/browse_step | a BrowseStepInput |
the step’s result: the accessibility tree, the text, the markdown, a screenshot’s hash, or a download set |
| si.act.capture | POST /v1/act/capture | the document, the network events, the address, the person’s identity | the capture with its source objects and the cells credited |
| si.act.crawl | POST /v1/act/crawl | a ScrapeRequest whose scope names the site’s seeds, with sitemap-driven discovery on |
the ScrapeRun, queued, or the run that already stands for the same template version and scope inside its freshness; every page lands as a capture through the one pipeline, and each page that could not be read is a gap with its reason |
| si.act.hold_identity | POST /v1/act/hold_identity | the platform, the scopes the rows declare | a browser session for the person; then the credential with its vault entry |
| si.act.probe_identity | POST /v1/act/probe_identity | the identity | its readiness with the reason |
| si.act.scrape_template | POST /v1/act/scrape_template | a TemplateInput |
the template, ready or not ready with the reason; for a preview, the rows from its samples and a bounded live scope |
| si.act.set_aside_run | POST /v1/act/set_aside_run | a SetAsideRunInput: the run’s key and a person’s plain reason |
the ScrapeRun, kept in the parked place with the reason, which the beat never takes up; each move is an event with the reason |
| si.act.sync_rule | POST /v1/act/sync_rule | the installed connector, the offer, the fields of section 2.8 | the rule; for a preview, what it would read and push |
| si.act.undo_action | POST /v1/act/undo_action | the move | the inverse move |
| si.act.watch_site | POST /v1/act/watch_site | a WatchSiteInput |
the watch, ready or not ready; for a preview, the pages it would read and the changes it would have raised over the window |
| si.work.backfill | POST /v1/work/backfill | a source, an access method, a window, a plane | the job; then partitions carrying the window |
| si.work.browse_open | POST /v1/work/browse_open | a BrowseOpenInput |
the session, opened, with its live view |
| si.work.browse_take_over | POST /v1/work/browse_take_over | a BrowseTakeOverInput |
the session, driven by the person or handed back |
| si.work.download_set | POST /v1/work/download_set | a DownloadInput |
DownloadJob rows, one per file |
| si.work.download_set_control | POST /v1/work/download_set_control | a DownloadControlInput |
the jobs in their new place |
The bars this system enforces#
Every bar is a row of the parameter register: an administrator changes it, and the platform reads it by name at the place of enforcement the row records.
| parameter | family | dimension | enforced at |
|---|---|---|---|
| the system’s declared capacity | machinery | jobs | dequeue |
| a run’s storage residency limit | machinery | bytes | every bulk-path write |
| attempt caps, by failure class | machinery | count | the reconciler |
| an access method’s politeness budget and rate limit | source | calls per unit of time | the call, under the lease |
| a browser’s page count and age before recycling | machinery | count; a time | the pool, at release |
| a browser pool’s leases open at once and the wait for one | machinery | count; a time | the pool, at each lease |
| the wait of a person’s capture for the record’s writer | machinery | a time (stipulated at ten seconds, under the extension’s wait of fifteen) | the capture tool, at the deposit: past the wait the capture is refused by name and sent again |
| a network proxy pool’s spend ceiling | receive | money per period | the exit, at the call |
| a browsing session’s longest life and its step rate | machinery | a time; steps per unit of time; milliseconds a step settles before its reading | the pool, at each step (the settle is 300 ms, stipulated, and a step that asks for no reading does not settle) |
| a download set’s residency and its per-file size cap | machinery | bytes | the download manager, at every block |
| the crawls a beat takes | machinery | runs per tick; runs open at once | the crawl beat, at each tick |
| the probe of a held identity | machinery | a time between two probes of one identity | the beat that probes the held identities, at each tick (six hours, stipulated) |
| an address found under a person’s held account | source | waits or runs | the crawl, as a walk queues what it found under a person’s own account: waits keeps it in the parked place with its reason |
Claims#
| claim | state | route or tool |
|---|---|---|
| The platform serves 28 calls of The acquisition system and the connector standard at contract 0476e35e6e275db5. | target | si.ask.get_content |