{"samples":[{"id":"6aae594c162093b9aff403af","input":"In the @divinci-ai/mcp client (McpClient), what is the default value of the `maxReconnectAttempts` configuration option?","target":"5 reconnect attempts before giving up."},{"id":"6aae594c162093b9aff403b1","input":"When a Cloudflare Worker proxies anonymous chat to a Divinci Release (the landing-page pattern), what shared secret authenticates the upstream calls instead of a Divinci API key?","target":"A shared HMAC secret, the `LANDING_PAGE_HMAC_KEY`."},{"id":"6aae594c162093b9aff403b3","input":"Which `window` global does the Divinci embed script (embed-script.js) expose by default, and which script-tag attribute renames it?","target":"It exposes `window.DIVINCI_AI` by default; set `data-global-name` to use a different name."},{"id":"6aae594c162093b9aff403b5","input":"What do the legacy `/sse` and `/message` paths of the Divinci hosted MCP server return?","target":"HTTP 410 Gone (they were retired 2026-08-07); clients must use the `/mcp` Streamable HTTP endpoint."},{"id":"6aae594c162093b9aff403b7","input":"In the @divinci-ai/client SDK, what is the default value of the `wsBaseUrl` (WebSocket origin) option on `DivinciClient`?","target":"`wss://live.divinci.app`."},{"id":"6aae594c162093b9aff403b9","input":"For a Divinci Release serving anonymous chat, what is the default value of `maxAnonymousChatMessages`?","target":"1 — it is a per-conversation message cap that you raise for multi-turn chat."},{"id":"6aae594c162093b9aff403bb","input":"What minimum Node.js version does the Divinci CLI (`@divinci-ai/cli`) require?","target":"Node.js 22.14.0 or later."},{"id":"6aae594c162093b9aff403bd","input":"Which Divinci workspace API key permission scope gates chat calls such as `chat.create` / `chat.send`?","target":"`transcript:write` — there is no separate chat scope, so an explicit permissions array without it gets 403 on chat calls."},{"id":"6aae594c162093b9aff403bf","input":"In the Divinci Server SDK's RAG API (`divinci.rag`), which vector store backs a collection by default?","target":"Qdrant; Pinecone serverless (`pinecone-serverless`) can be chosen instead."},{"id":"6aae594c162093b9aff403c1","input":"Of the 33 providers in Divinci's BYOK (bring-your-own-key) system, which two do not accept a workspace BYOK credential?","target":"`divinci` and `redis`, because they are platform infrastructure."},{"id":"6aae594c162093b9aff403c3","input":"For Divinci hosted Hermes agents, what is the default per-workspace cap on agents, and when does an agent's container first spin up?","target":"The default cap is 10 agents per workspace; creation is lazy, so no container spins until the agent's first chat."},{"id":"6aae594c162093b9aff403c5","input":"On a Divinci Release, what is the maximum length of each conversation starter (the clickable prompts on an empty chat)?","target":"200 characters each."},{"id":"6aae594c162093b9aff403c7","input":"What happens if you set `sttToolOverride` or `pricingTiers` on a Divinci Release while pinned to an older version of `@divinci-ai/server`?","target":"The fields are dropped silently rather than raising an error, so you must upgrade the SDK before relying on them."},{"id":"6aae594c162093b9aff403c9","input":"What is the default value of a Divinci release's `localModel` policy for on-device (browser-extension) inference, and how does it behave?","target":"`fallback`: use the device first when the extension is present and the visitor's toggle is on, and fall back to the cloud release on any local error."},{"id":"6aae594c162093b9aff403cb","input":"What does the Divinci embed script's `read-aloud` attribute use to speak assistant messages?","target":"The browser's own `speechSynthesis` — no server TTS and no cost, and it works for anonymous visitors."},{"id":"6aae594c162093b9aff403cd","input":"Which kinds of failures cause a Divinci Release to fall through to its ordered `fallbackAssistants` chain?","target":"Retryable (transient) failures: a provider 429, a 5xx, a capacity or quota error, or a timeout. Refusals, malformed prompts and content-filter blocks do not fail over."},{"id":"6aae594c162093b9aff403cf","input":"For a Divinci RAG vector group attached to a Release, what are the default `mergeStrategy` and `maxChunksPerVector`?","target":"`mergeStrategy` defaults to `interleave` and `maxChunksPerVector` defaults to 3."},{"id":"6aae594c162093b9aff403d1","input":"When creating a vector index with `client.rag.vectorIndex.create` in the @divinci-ai/client SDK, which `vectorIndexTool` value is the managed default backend?","target":"`vectorize`."},{"id":"6aae594c162093b9aff403d3","input":"What error does the Divinci chat message endpoint return if a request omits both `assistantName` and `replyTo`?","target":"`400 bad form data`."},{"id":"6aae594c162093b9aff403d5","input":"Does `client.chat.stream()` in the @divinci-ai/client SDK return an async iterator that can be consumed with `for await`?","target":"No. Deltas arrive through the `onChunk` / `onEvent` callbacks, and the returned promise resolves once with the complete message."},{"id":"6aae594c162093b9aff403d7","input":"On the Divinci free plan, which models can be used without attaching a BYOK key?","target":"Google-authored and Cloudflare-hosted models; others such as OpenAI, Anthropic, Mistral and Moonshot/Kimi are BYOK-only on the free plan."},{"id":"6aae594c162093b9aff403d9","input":"When a visitor clicks a link-out Divinci Chat CTA card, what parameter is appended to the CTA's `targetUrl`?","target":"A `ref=divinci` parameter (the URL opens in a new tab)."},{"id":"6aae594c162093b9aff403db","input":"Which @divinci-ai/client error class is thrown with HTTP status 402?","target":"`PaymentRequiredError`, thrown when an x402-priced resource needs payment."},{"id":"6aae594c162093b9aff403dd","input":"How is voice-call usage rounded when Divinci bills a phone call handled by a Release?","target":"Each call's duration is rounded up to the whole minute, so a 61-second call bills as 2 minutes."},{"id":"6aae594c162093b9aff403df","input":"When enabling automatic re-sync of an e-commerce product catalog in Divinci's Feed Digester, which intervals can be chosen?","target":"6 hours, 12 hours, daily, or weekly."},{"id":"6aae594c162093b9aff403e1","input":"Does the @divinci-ai/client SDK automatically refresh a `userToken` passed to `DivinciClient`?","target":"No. It is sent verbatim as the Bearer value; the client has no built-in refresh logic, so it must be a short-lived access token."},{"id":"6aae594c162093b9aff403e3","input":"When setting up the open-source Divinci demo pipeline with the Divinci CLI, why must `DIVINCI_API_KEY` be left unset?","target":"Workspace creation needs the OAuth session, and the CLI prefers an API key over OAuth whenever one is present, so a key in the shell breaks `divinci workspace create`."},{"id":"6aae594c162093b9aff403e5","input":"In the Divinci Docs workflow, what does the `--swap` flag do on `divinci connectors web-crawl`?","target":"The new crawl is built alongside the live vector and swapped in only on success, so a failed re-crawl leaves the previous corpus in place."},{"id":"6aae594c162093b9aff403e7","input":"How many probes and attack classes make up Divinci's red-team corpus, and how many probes at most go into one billed request?","target":"32 probes in seven classes; each class is one billed request of at most six probes."},{"id":"6aae594c162093b9aff403e9","input":"Which event does the Divinci MCP server push when its resource catalog changes?","target":"`notifications/resources/list_changed`, which the @divinci-ai/mcp client surfaces via `onResourcesChanged`."},{"id":"6aae594c162093b9aff403eb","input":"In the @divinci-ai/server SDK's x402 payments, what does the wallet sign when a call returns 402?","target":"An EIP-3009 `transferWithAuthorization` for the exact amount required; it does not broadcast, and the request is retried with the `X-PAYMENT` header."},{"id":"6aae594c162093b9aff403ed","input":"What is the default `DAILY_GLOBAL_USD` limit in the Divinci Server SDK's Spend Guard?","target":"$50 per day platform-wide (a development default)."},{"id":"6aae594c162093b9aff403ef","input":"Where does the @divinci-ai/mcp client keep Auth0 tokens by default?","target":"In memory, so they are lost on page reload; supply a custom `TokenStorage` to persist them."},{"id":"6aae594c162093b9aff403f1","input":"When publishing a Divinci release as its own MCP server, what must be configured before `allowAnonymousMcp` can be enabled?","target":"A positive `maxSpendPerTokenCents` spend cap; anonymous access without a spend cap is rejected."},{"id":"6aae594d162093b9aff403f3","input":"Which `divinci hermes update` flag removes a hosted Hermes agent's BYOK provider key and reverts it to platform billing?","target":"`--clear-provider-key`."},{"id":"6aae594d162093b9aff403f5","input":"What happens if a Divinci release update body includes the old `toolRouting` object?","target":"It still returns 200 and changes nothing, because tool routing was merged into `skillConfig` and the update route ignores unknown keys."},{"id":"6aae594d162093b9aff403f7","input":"Which workspace setting restricts which origins can load an embedded Divinci Release?","target":"`allowedDomains`; an unset allowlist is not origin-restricted."},{"id":"6aae594d162093b9aff403f9","input":"Is Llama Guard harmful-message moderation on or off by default for a Divinci Release?","target":"On by default (`noHarmful.on: true`) — it is opt-out, unlike the other moderation controls."},{"id":"6aae594d162093b9aff403fb","input":"Which of Divinci's three caching layers skips the entire turn, including RAG, moderation, escrow and the LLM call?","target":"The Redis response-level semantic cache (off by default, anonymous/public chat only)."},{"id":"6aae594d162093b9aff403fd","input":"When de-identification is enabled on a Divinci Release, is the original un-redacted message text archived?","target":"No — it is discarded, not archived, so a redacted message cannot be un-redacted later."},{"id":"6aae594d162093b9aff403ff","input":"In Divinci Scored QA calibration, which statistics measure agreement between human raters and LLM judges?","target":"Spearman's rho and Quadratic Weighted Kappa; the highest-correlation judge becomes the workspace default."},{"id":"6aae594d162093b9aff40401","input":"In the Divinci Server SDK's Site Converter (`divinci.siteToCode.createAndWait`), what is the maximum number of `pageUrls` that can be converted?","target":"20."},{"id":"6aae594d162093b9aff40403","input":"What is the default polling timeout for `divinci trust run` (TrustBench) in the Divinci CLI?","target":"600 seconds."},{"id":"6aae594d162093b9aff40405","input":"In Divinci's web scraping cascade, which tool is tried first as the cheapest option?","target":"`@divinci-ai/fetch-scraper`, a plain HTTP fetch."},{"id":"6aae594d162093b9aff40407","input":"Why is Divinci's `@divinci-ai/openparse` parser deprecated, and what should be used instead?","target":"It hit memory limits on large PDFs; LangExtract replaces it."},{"id":"6aae594d162093b9aff40409","input":"In Divinci's bring-your-own retriever contract, what range must each result's `score` fall in?","target":"0–1; unbounded scores such as BM25 must be squashed monotonically before returning."},{"id":"6aae594d162093b9aff4040b","input":"Which Divinci embedding model is recommended for catalog imports and batches of more than ~100 files?","target":"`gemini-embedding-001@1536` (~100K requests/min, batched)."},{"id":"6aae594d162093b9aff4040d","input":"Which base models can be fine-tuned through the Cloudflare Workers AI LoRA backend in Divinci?","target":"Gemma 3 12B IT and Mistral Small 3.1 24B Instruct."},{"id":"6aae594d162093b9aff4040f","input":"How does Divinci's `stanford` de-identification engine handle the `hybrid` strategy?","target":"It treats `hybrid` as plain `redact` (only the default `presidio` engine implements the real mix)."},{"id":"6aae594d162093b9aff40411","input":"In `@divinci-ai/extension-bridge`, what does `chat()` do when the Divinci Local Inference extension is not installed?","target":"It falls back to the supplied `cloudChat`, or rejects if no `cloudChat` was supplied."},{"id":"6aae594d162093b9aff40413","input":"Which post types does the Divinci WordPress connector (`divinci connectors wordpress`) ingest by default?","target":"`posts,pages` (set other `rest_base` values with `--post-types`)."},{"id":"6aae594d162093b9aff40415","input":"What is the default page limit for `divinci connectors web-crawl`?","target":"80 pages (`--limit`)."},{"id":"6aae594d162093b9aff40417","input":"Do a Divinci API key's `transcript:read` / `transcript:write` scopes grant access to tools over the Divinci MCP server (e.g. from n8n or Make)?","target":"No. They are not MCP scope names, so they contribute no MCP access even though they are the chat scopes for the REST API."},{"id":"6aae594d162093b9aff40419","input":"What happens if you trigger a Divinci connector sync while a sync job for that connection is already running?","target":"You get a 409 with the running job's id."},{"id":"6aae594d162093b9aff4041b","input":"What is the default `dailyRequestBudget` for the Divinci Practice Better connector?","target":"2000 API calls per day."},{"id":"6aae594d162093b9aff4041d","input":"When does Divinci renew a Google Drive connector's `changes.watch` push channel?","target":"Within 12 hours of it lapsing; if registration or renewal fails, the connection falls back to polling."},{"id":"6aae594d162093b9aff4041f","input":"For the Divinci Cloudflare R2 connector, what value does the `endpoint` credential field expect?","target":"Your Cloudflare account ID, not a URL."},{"id":"6aae594d162093b9aff40421","input":"Per the Divinci changelog of August 11, 2026, what is the daily file upload cap on the free and premium plans?","target":"50 files per day on free and 100 on premium, now also applied to bulk uploads."},{"id":"6aae594d162093b9aff40423","input":"In Divinci, what happens to a chat turn in a HIPAA-scoped workspace if the model provider is not covered by a business associate agreement?","target":"The turn is refused; HIPAA-scoped protections cannot be switched off once enabled."},{"id":"6aae594d162093b9aff40425","input":"Per the Divinci changelog (August 12, 2026), what intervals can a hosted Hermes agent's proactive loop run on?","target":"Every 3 hours, every 6, up to every 24 hours, instead of a single daily slot."}],"scorerKey":"llm-factual-consistency-vs-reference"}