AI & research · One purpose, multiple implementations
AI Document Assistant via Telegram + Supabase
Adapt the published “AI Document Assistant via Telegram + Supabase” workflow to other implementations while preserving its documented behavior.
The source contains 25 executable nodes and 26 connections. This draft is derived from the n8n export; it needs a human-reviewed purpose, input/output contract, and behavioral tests before a port can claim equivalence.
Open agent files on GitHub ↗Shared workflow
This initial Mermaid diagram shows observed connections from the n8n source, including model and tool dependencies. Unresolved connections are listed under Evidence. It is a starting point for the shared design; it does not execute the agent or establish correctness.
Rendering the workflow diagram…
Read the Mermaid source
flowchart LR n0["Google Gemini Chat Model"] n1["OpenWeatherMap"] n2["AI Agent"] n3["Telegram Trigger"] n4["Telegram"] n5["Think"] n6["Embeddings Google Gemini"] n7["Default Data Loader"] n8["Recursive Character Text Splitter"] n9["Extract from File"] n10["Answer questions with a vector store"] n11["Supabase Vector Store"] n12["Telegram - Embedding Complete"] n13["Supabase - Save Embeddings"] n14["Command Router"] n15["Telegram - Download file"] n16["Aggregate"] n17["Fallback- No formatting"] n18["Split Out"] n19["Simple Memory"] n20["Handle formatting and split"] n21["Manual Mapping"] n22["Unsupported message"] n23["Send processing document message"] n24["Send embedding Started message"] n5 -->|"ai_tool out 0 to in 0"| n2 n2 -->|"main out 0 to in 0"| n20 n4 -->|"main out 1 to in 0"| n17 n16 -->|"main out 0 to in 0"| n12 n18 -->|"main out 0 to in 0"| n21 n19 -->|"ai_memory out 0 to in 0"| n2 n14 -->|"main out 0 to in 0"| n15 n14 -->|"main out 1 to in 0"| n2 n14 -->|"main out 2 to in 0"| n22 n21 -->|"main out 0 to in 0"| n4 n1 -->|"ai_tool out 0 to in 0"| n2 n3 -->|"main out 0 to in 0"| n14 n9 -->|"main out 0 to in 0"| n13 n9 -->|"main out 0 to in 0"| n24 n7 -->|"ai_document out 0 to in 0"| n13 n11 -->|"ai_vectorStore out 0 to in 0"| n10 n6 -->|"ai_embedding out 0 to in 0"| n13 n6 -->|"ai_embedding out 0 to in 0"| n11 n0 -->|"ai_languageModel out 0 to in 0"| n2 n0 -->|"ai_languageModel out 0 to in 0"| n10 n15 -->|"main out 0 to in 0"| n9 n15 -->|"main out 0 to in 0"| n23 n13 -->|"main out 0 to in 0"| n16 n20 -->|"main out 0 to in 0"| n18 n8 -->|"ai_textSplitter out 0 to in 0"| n7 n10 -->|"ai_tool out 0 to in 0"| n2
Implementations
Each implementation shares the agent’s purpose. Available source, planned ports, and verified results are labeled separately.
n8n
Source available · not runtime-tested
Read setup and source ↗Inspect the n8n canvas and download JSON →Build or operate with an assistant
Each prompt includes separate build and operating instructions tailored to this agent. They are untested instructions, not working implementations. Check the chosen product’s actual capabilities before using it.
Evidence and remaining work
The bulk collection has publication checks, not live-service testing. Diagrams and prompts are documentation. A port needs reviewed schemas, fixtures, actual execution results, and explicit limits before claiming equivalent behavior.
- Define input/output schemas and expected outputs.
- Review source expressions, external actions, and failure behavior.
- Determine whether model judgment is required or deterministic rules suffice.
Curated by Nik Bear Brown. Preserve each implementation’s original attribution and license. These source templates are separate from the audited Claude directory.