[{"data":1,"prerenderedAt":466},["ShallowReactive",2],{"docs-\u002Fdocs\u002Fai-chat":3},{"id":4,"title":5,"body":6,"description":457,"extension":458,"meta":459,"navigation":460,"path":461,"redirect":462,"seo":463,"stem":464,"__hash__":465},"docs\u002Fdocs\u002Fai-chat\u002Findex.md","AI Chat",{"type":7,"value":8,"toc":440},"minimark",[9,13,28,35,47,52,79,87,92,104,142,149,153,156,160,172,192,196,203,261,265,272,290,302,306,309,342,346,352,356,408,415,419],[10,11,5],"h1",{"id":12},"ai-chat",[14,15,16,19,20,23,24,27],"p",{},[17,18,5],"strong",{}," is the AI assistant built into the DBConvert Streams desktop app.\nOpen it, pick an agent, and ask about your databases, files, and streams in\nplain language. It works from your ",[17,21,22],{},"live"," workspace — the same connections,\nschemas, files, and stream status you see in Data Explorer — and it is\n",[17,25,26],{},"read-only",": it can look and advise, but never changes your data.",[14,29,30,31,34],{},"It runs your ",[17,32,33],{},"own"," agent (Claude Code, Codex, GitHub Copilot CLI, or OpenCode)\nwith your own model key — DBConvert injects its tools into the session\nautomatically, so there's no separate setup.",[36,37,38],"blockquote",{},[14,39,40,41,46],{},"Want to use DBConvert from an external AI app (Claude Desktop, Cursor) or over\nHTTP instead? See ",[42,43,45],"a",{"href":44},"\u002Fdocs\u002Fmcp","AI Clients (MCP)",".",[48,49,51],"h2",{"id":50},"start-a-chat","Start a chat",[53,54,55,62,68,71],"ol",{},[56,57,58,59,46],"li",{},"Open the DBConvert Streams ",[17,60,61],{},"desktop app",[56,63,64,65,67],{},"Click ",[17,66,5],{}," in the left sidebar.",[56,69,70],{},"Pick an installed agent, and start a session.",[56,72,73,74,78],{},"Ask: ",[75,76,77],"code",{},"List my DBConvert connections."," — if the connections you see in Data\nExplorer come back, you're ready.",[36,80,81],{},[14,82,83,86],{},[17,84,85],{},"No agent installed?"," AI Chat runs an agent you already have on your machine.\nInstall one of the supported agent CLIs and it appears in the picker.",[88,89,91],"h3",{"id":90},"ask-ai-in-context","Ask AI in context",[14,93,94,95,98,99,103],{},"You don't have to start from a blank chat. ",[17,96,97],{},"Ask AI"," buttons across the app open\na chat already scoped to what you're looking at, so the assistant answers about\n",[100,101,102],"em",{},"that"," thing without enumerating your whole workspace first:",[105,106,107,114,120,130,136],"ul",{},[56,108,109,110,113],{},"a ",[17,111,112],{},"table or view"," (Data \u002F Structure \u002F Summary),",[56,115,109,116,119],{},[17,117,118],{},"file"," (CSV \u002F JSON \u002F Parquet),",[56,121,109,122,125,126,129],{},[17,123,124],{},"connection"," or ",[17,127,128],{},"database",",",[56,131,109,132,135],{},[17,133,134],{},"SQL console"," (including a multi-source federated one),",[56,137,109,138,141],{},[17,139,140],{},"stream"," config (from its Monitor \u002F Compare \u002F History tabs).",[14,143,144,145,148],{},"Each scope also offers a few one-click ",[17,146,147],{},"suggestions"," tuned to it — e.g. \"Find\ndata quality issues in this table\", \"Show the first rows\" for a file, or, in a\nmulti-source console, \"Compare a table with a file\".",[48,150,152],{"id":151},"what-you-can-ask","What you can ask",[14,154,155],{},"Real prompts, real answers — every exchange below came back from a live\nworkspace, numbers and all.",[88,157,159],{"id":158},"database-inspection","Database inspection",[36,161,162],{},[14,163,164,167,168,171],{},[17,165,166],{},"You:"," Describe the ",[75,169,170],{},"orders"," table on my Postgres demo connection — columns, keys, indexes. Lay it out as a table.",[36,173,174],{},[14,175,176,179,180,183,184,187,188,191],{},[17,177,178],{},"AI:"," ",[75,181,182],{},"public.orders"," — 10 columns, primary key ",[75,185,186],{},"order_id"," … No index on ",[75,189,190],{},"region"," — add one if you group orders by region.",[88,193,195],{"id":194},"analytics-real-numbers-not-a-guess","Analytics — real numbers, not a guess",[36,197,198],{},[14,199,200,202],{},[17,201,166],{}," What are my top film categories by rental revenue? (sakila sample)",[36,204,205,209,254],{},[14,206,207],{},[17,208,178],{},[210,211,212,226],"table",{},[213,214,215],"thead",{},[216,217,218,222],"tr",{},[219,220,221],"th",{},"Category",[219,223,225],{"align":224},"right","Revenue",[227,228,229,238,246],"tbody",{},[216,230,231,235],{},[232,233,234],"td",{},"Sports",[232,236,237],{"align":224},"$5,314.21",[216,239,240,243],{},[232,241,242],{},"Sci-Fi",[232,244,245],{"align":224},"$4,756.98",[216,247,248,251],{},[232,249,250],{},"Animation",[232,252,253],{"align":224},"$4,656.30",[14,255,256,257,260],{},"Joined ",[75,258,259],{},"category → film_category → film → inventory → rental → payment",", ran in ~55 ms. Want the query to reuse?",[88,262,264],{"id":263},"stream-troubleshooting","Stream troubleshooting",[36,266,267],{},[14,268,269,271],{},[17,270,166],{}," My MySQL → PostgreSQL CDC stream stopped overnight. List my streams, check status and recent errors, and tell me what to fix.",[36,273,274],{},[14,275,276,278,279,179,282,285,286,289],{},[17,277,178],{}," Found it — the latest run of ",[75,280,281],{},"10M-cdc-mysql2pg",[17,283,284],{},"FAILED"," with ",[75,287,288],{},"Unknown database 'target'",", source-side. The CDC source points at a database that doesn't exist on the MySQL server. Fix: create it (or correct the name on the connection), then restart.",[14,291,292,293,297,298,46],{},"More complete transcripts: ",[42,294,296],{"href":295},"\u002Fdocs\u002Fmcp\u002Fexamples\u002Fsql-exploration","plain-English SQL exploration","\nand ",[42,299,301],{"href":300},"\u002Fdocs\u002Fmcp\u002Fexamples\u002Ffederated-validation","federated Postgres + Parquet validation",[48,303,305],{"id":304},"files-and-multi-source","Files and multi-source",[14,307,308],{},"AI Chat isn't limited to databases.",[105,310,311,320],{},[56,312,313,316,317,46],{},[17,314,315],{},"Files."," For a CSV \u002F JSON \u002F Parquet file, the chat reads its schema,\npreviews rows, and queries it via DuckDB — no hand-written ",[75,318,319],{},"read_csv",[56,321,322,325,326,329,330,333,334,337,338,341],{},[17,323,324],{},"Multi-source (federated)."," In a SQL console with two or more sources\nattached — say a MySQL database and a folder of CSVs — AI Chat sees ",[17,327,328],{},"every","\nsource and its alias. Ask it to ",[100,331,332],{},"join",", ",[100,335,336],{},"find the join keys",", or ",[100,339,340],{},"compare and\nreconcile"," across them, and it writes the cross-source query and runs it. This\nis the everyday \"is my CSV export still in sync with the live table?\" check,\ndone in one question.",[48,343,345],{"id":344},"generate-sql","Generate SQL",[14,347,348,349,351],{},"Next to any SQL console you'll also find ",[17,350,345],{},": type a plain-language\nrequest and it writes the SQL straight into the editor (including federated\nqueries across multiple sources), for you to review and run. It's the same\nengine as AI Chat, aimed at the editor instead of a conversation.",[48,353,355],{"id":354},"privacy-safety","Privacy & safety",[105,357,358,396,402],{},[56,359,360,363,364,367,368,371,372,367,375,367,378,381,382,367,385,367,388,391,392,395],{},[17,361,362],{},"Read-only."," Only ",[75,365,366],{},"SELECT"," \u002F ",[75,369,370],{},"WITH … SELECT"," reach your database; everything\nelse is rejected before it runs. No ",[75,373,374],{},"INSERT",[75,376,377],{},"UPDATE",[75,379,380],{},"DELETE",", no\n",[75,383,384],{},"DROP",[75,386,387],{},"ALTER",[75,389,390],{},"CREATE",", no connection or stream changes. When the AI\nsuggests a change, ",[17,393,394],{},"you"," run it.",[56,397,398,401],{},[17,399,400],{},"Your keys, your model."," The agent uses your own account; your model key\nnever passes through DBConvert.",[56,403,404,407],{},[17,405,406],{},"Credentials never leave."," Tools return schemas, rows, and stream state —\nnever passwords or cloud keys. Tool results and sample data do reach your\nmodel provider (that's how any AI assistant works); nothing is sent anywhere on\nits own.",[14,409,410,411,46],{},"Full boundary: ",[42,412,414],{"href":413},"\u002Fdocs\u002Fmcp\u002Fsafety-and-privacy","Safety & privacy",[48,416,418],{"id":417},"related","Related",[105,420,421,426,433],{},[56,422,423,425],{},[42,424,45],{"href":44}," — use DBConvert from an external AI app or over HTTP.",[56,427,428,432],{},[42,429,431],{"href":430},"\u002Fdocs\u002Fmcp\u002Ftools","Tools reference"," — every read-only tool the assistant can call.",[56,434,435,439],{},[42,436,438],{"href":437},"\u002Fdocs\u002Fdatabase-explorer","Data Explorer"," — the workspace AI Chat reads from.",{"title":441,"searchDepth":442,"depth":442,"links":443},"",2,[444,448,453,454,455,456],{"id":50,"depth":442,"text":51,"children":445},[446],{"id":90,"depth":447,"text":91},3,{"id":151,"depth":442,"text":152,"children":449},[450,451,452],{"id":158,"depth":447,"text":159},{"id":194,"depth":447,"text":195},{"id":263,"depth":447,"text":264},{"id":304,"depth":442,"text":305},{"id":344,"depth":442,"text":345},{"id":354,"depth":442,"text":355},{"id":417,"depth":442,"text":418},"The built-in AI assistant inside DBConvert Streams. Ask about your databases, files, and streams in plain language — it reads your live workspace and never writes.","md",{},false,"\u002Fdocs\u002Fai-chat",null,{"title":5,"description":457},"docs\u002Fai-chat\u002Findex","k1j-C36bmWggpBvcyAIy28Y9XCXFyfWxnkdm0yQt5lM",1783896283069]