21 KiB
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
AI agent with Telegram bot interface and SvelteKit web frontend, using Convex ( self-hosted) as the backend database.
Architecture
stealth-ai-relay/
├── backend/ # Python Telegram bot (aiogram + pydantic-ai)
├── frontend/ # SvelteKit + Svelte 5 + Tailwind CSS 4 + Convex
└── caddy/ # Reverse proxy config
- Convex is self-hosted - all env vars are configured, use
CONVEX_SELF_HOSTED_URLfor backend /apipath is reserved - use/servicefor internal HTTP endpoints in frontend- Frontend lives at root, Convex at
/convex, Convex HTTP at/convex-http - No code comments - do not write comments in code, code should be self-documenting
Commands
Backend (Python)
cd backend
# Type checking
ty check
# Linting and formatting
ruff format
ruff check --fix
Frontend (SvelteKit)
cd frontend
bun check # Type check
bun format # Format with Prettier
bun lint # Lint with ESLint + Prettier
MCP Tools
Context7
Use mcp__context7__resolve-library-id then mcp__context7__query-docs to fetch
up-to-date documentation for any library (Convex, aiogram, pydantic-ai, etc.).
Svelte MCP (REQUIRED for Svelte work)
mcp__svelte__list-sections- discover available docsmcp__svelte__get-documentation- fetch relevant sectionsmcp__svelte__svelte-autofixer- MUST use before sending any Svelte codemcp__svelte__playground-link- generate playground links (ask user first)
Svelte 5 + SvelteKit
Setup Convex in SvelteKit
<!-- src/routes/+layout.svelte -->
<script lang="ts">
import {PUBLIC_CONVEX_URL} from '$env/static/public';
import {setupConvex} from 'convex-svelte';
const {children} = $props();
setupConvex(PUBLIC_CONVEX_URL);
</script>
{@render children()}
Using Convex queries/mutations
<script lang="ts">
import {useQuery, useMutation} from 'convex-svelte';
import {api} from '$lib/convex/_generated/api.js';
const tasks = useQuery(api.tasks.list, {});
const createTask = useMutation(api.tasks.create);
</script>
{#if tasks.isLoading}
Loading...
{:else if tasks.error}
Error: {tasks.error.toString()}
{:else}
{#each tasks.data as task}
<div>{task.text}</div>
{/each}
{/if}
Python Backend
Uses aiogram 3.x for Telegram bot, pydantic-settings for config, pydantic-ai for AI features.
Logging pattern
from utils.logging import logger
logger.info("Test")
Svelte
You are able to use the Svelte MCP server, where you have access to comprehensive Svelte 5 and SvelteKit documentation. Here's how to use the available tools effectively:
Available MCP Tools:
1. list-sections
Use this FIRST to discover all available documentation sections. Returns a structured list with titles, use_cases, and paths. When asked about Svelte or SvelteKit topics, ALWAYS use this tool at the start of the chat to find relevant sections.
2. get-documentation
Retrieves full documentation content for specific sections. Accepts single or multiple sections. After calling the list-sections tool, you MUST analyze the returned documentation sections (especially the use_cases field) and then use the get-documentation tool to fetch ALL documentation sections that are relevant for the user's task.
3. svelte-autofixer
Analyzes Svelte code and returns issues and suggestions. You MUST use this tool whenever writing Svelte code before sending it to the user. Keep calling it until no issues or suggestions are returned.
4. playground-link
Generates a Svelte Playground link with the provided code. After completing the code, ask the user if they want a playground link. Only call this tool after user confirmation and NEVER if code was written to files in their project.
Convex guidelines
Function guidelines
New function syntax
- ALWAYS use the new function syntax for Convex functions. For example:
import {query} from "./_generated/server";
import {v} from "convex/values";
export const f = query({
args: {},
returns: v.null(),
handler: async (ctx, args) => {
// Function body
},
});
Http endpoint syntax
- HTTP endpoints are defined in
convex/http.tsand require anhttpActiondecorator. For example:
import {httpRouter} from "convex/server";
import {httpAction} from "./_generated/server";
const http = httpRouter();
http.route({
path: "/echo",
method: "POST",
handler: httpAction(async (ctx, req) => {
const body = await req.bytes();
return new Response(body, {status: 200});
}),
});
- HTTP endpoints are always registered at the exact path you specify in the
pathfield. For example, if you specify/api/someRoute, the endpoint will be registered at/api/someRoute.
Validators
- Below is an example of an array validator:
import {mutation} from "./_generated/server";
import {v} from "convex/values";
export default mutation({
args: {
simpleArray: v.array(v.union(v.string(), v.number())),
},
handler: async (ctx, args) => {
//...
},
});
- Below is an example of a schema with validators that codify a discriminated union type:
import {defineSchema, defineTable} from "convex/server";
import {v} from "convex/values";
export default defineSchema({
results: defineTable(
v.union(
v.object({
kind: v.literal("error"),
errorMessage: v.string(),
}),
v.object({
kind: v.literal("success"),
value: v.number(),
}),
),
)
});
- Always use the
v.null()validator when returning a null value. Below is an example query that returns a null value:
import {query} from "./_generated/server";
import {v} from "convex/values";
export const exampleQuery = query({
args: {},
returns: v.null(),
handler: async (ctx, args) => {
console.log("This query returns a null value");
return null;
},
});
- Here are the valid Convex types along with their respective validators:
Convex Type TS/JS type Example Usage Validator for argument validation and schemas Notes Id string doc._idv.id(tableName)Null null nullv.null()JavaScript's undefinedis not a valid Convex value. Functions the returnundefinedor do not return will returnnullwhen called from a client. Usenullinstead.Int64 bigint 3nv.int64()Int64s only support BigInts between -2^63 and 2^63-1. Convex supports bigints in most modern browsers.Float64 number 3.1v.number()Convex supports all IEEE-754 double-precision floating point numbers (such as NaNs). Inf and NaN are JSON serialized as strings. Boolean boolean truev.boolean()String string "abc"v.string()Strings are stored as UTF-8 and must be valid Unicode sequences. Strings must be smaller than the 1MB total size limit when encoded as UTF-8. Bytes ArrayBuffer new ArrayBuffer(8)v.bytes()Convex supports first class bytestrings, passed in as ArrayBuffers. Bytestrings must be smaller than the 1MB total size limit for Convex types.Array Array [1, 3.2, "abc"]v.array(values)Arrays can have at most 8192 values. Object Object {a: "abc"}v.object({property: value})Convex only supports "plain old JavaScript objects" (objects that do not have a custom prototype). Objects can have at most 1024 entries. Field names must be nonempty and not start with "$" or "_". Record Record {"a": "1", "b": "2"}v.record(keys, values)Records are objects at runtime, but can have dynamic keys. Keys must be only ASCII characters, nonempty, and not start with "$" or "_".
Function registration
- Use
internalQuery,internalMutation, andinternalActionto register internal functions. These functions are private and aren't part of an app's API. They can only be called by other Convex functions. These functions are always imported from./_generated/server. - Use
query,mutation, andactionto register public functions. These functions are part of the public API and are exposed to the public Internet. Do NOT usequery,mutation, oractionto register sensitive internal functions that should be kept private. - You CANNOT register a function through the
apiorinternalobjects. - ALWAYS include argument and return validators for all Convex functions. This includes
all of
query,internalQuery,mutation,internalMutation,action, andinternalAction. If a function doesn't return anything, includereturns: v.null()as its output validator. - If the JavaScript implementation of a Convex function doesn't have a return value, it
implicitly returns
null.
Function calling
- Use
ctx.runQueryto call a query from a query, mutation, or action. - Use
ctx.runMutationto call a mutation from a mutation or action. - Use
ctx.runActionto call an action from an action. - ONLY call an action from another action if you need to cross runtimes (e.g. from V8 to Node). Otherwise, pull out the shared code into a helper async function and call that directly instead.
- Try to use as few calls from actions to queries and mutations as possible. Queries and mutations are transactions, so splitting logic up into multiple calls introduces the risk of race conditions.
- All of these calls take in a
FunctionReference. Do NOT try to pass the callee function directly into one of these calls. - When using
ctx.runQuery,ctx.runMutation, orctx.runActionto call a function in the same file, specify a type annotation on the return value to work around TypeScript circularity limitations. For example,
export const f = query({
args: { name: v.string() },
returns: v.string(),
handler: async (ctx, args) => {
return "Hello " + args.name;
},
});
export const g = query({
args: {},
returns: v.null(),
handler: async (ctx, args) => {
const result: string = await ctx.runQuery(api.example.f, { name: "Bob" });
return null;
},
});
Function references
- Function references are pointers to registered Convex functions.
- Use the
apiobject defined by the framework inconvex/_generated/api.tsto call public functions registered withquery,mutation, oraction. - Use the
internalobject defined by the framework inconvex/_generated/api.tsto call internal (or private) functions registered withinternalQuery,internalMutation, orinternalAction. - Convex uses file-based routing, so a public function defined in
convex/example.tsnamedfhas a function reference ofapi.example.f. - A private function defined in
convex/example.tsnamedghas a function reference ofinternal.example.g. - Functions can also registered within directories nested within the
convex/folder. For example, a public functionhdefined inconvex/messages/access.tshas a function reference ofapi.messages.access.h.
Api design
- Convex uses file-based routing, so thoughtfully organize files with public query,
mutation, or action functions within the
convex/directory. - Use
query,mutation, andactionto define public functions. - Use
internalQuery,internalMutation, andinternalActionto define private, internal functions.
Pagination
- Paginated queries are queries that return a list of results in incremental pages.
- You can define pagination using the following syntax:
import {v} from "convex/values";
import {query, mutation} from "./_generated/server";
import {paginationOptsValidator} from "convex/server";
export const listWithExtraArg = query({
args: {paginationOpts: paginationOptsValidator, author: v.string()},
handler: async (ctx, args) => {
return await ctx.db
.query("messages")
.filter((q) => q.eq(q.field("author"), args.author))
.order("desc")
.paginate(args.paginationOpts);
},
});
Note: paginationOpts is an object with the following properties:
numItems: the maximum number of documents to return (the validator isv.number())cursor: the cursor to use to fetch the next page of documents (the validator isv.union(v.string(), v.null()))- A query that ends in
.paginate()returns an object that has the following properties:- page (contains an array of documents that you fetches)
- isDone (a boolean that represents whether or not this is the last page of documents)
- continueCursor (a string that represents the cursor to use to fetch the next page of documents)
Validator guidelines
v.bigint()is deprecated for representing signed 64-bit integers. Usev.int64()instead.- Use
v.record()for defining a record type.v.map()andv.set()are not supported.
Schema guidelines
- Always define your schema in
convex/schema.ts. - Always import the schema definition functions from
convex/server. - System fields are automatically added to all documents and are prefixed with an
underscore. The two system fields that are automatically added to all documents are
_creationTimewhich has the validatorv.number()and_idwhich has the validatorv.id(tableName). - Always include all index fields in the index name. For example, if an index is defined
as
["field1", "field2"], the index name should be "by_field1_and_field2". - Index fields must be queried in the same order they are defined. If you want to be able to query by "field1" then "field2" and by "field2" then "field1", you must create separate indexes.
Typescript guidelines
- You can use the helper typescript type
Idimported from './_generated/dataModel' to get the type of the id for a given table. For example if there is a table called ' users' you can useId<'users'>to get the type of the id for that table. - If you need to define a
Recordmake sure that you correctly provide the type of the key and value in the type. For example a validatorv.record(v.id('users'), v.string())would have the typeRecord<Id<'users'>, string>. Below is an example of usingRecordwith anIdtype in a query:
import {query} from "./_generated/server";
import {Doc, Id} from "./_generated/dataModel";
export const exampleQuery = query({
args: {userIds: v.array(v.id("users"))},
returns: v.record(v.id("users"), v.string()),
handler: async (ctx, args) => {
const idToUsername: Record<Id<"users">, string> = {};
for (const userId of args.userIds) {
const user = await ctx.db.get("users", userId);
if (user) {
idToUsername[user._id] = user.username;
}
}
return idToUsername;
},
});
- Be strict with types, particularly around id's of documents. For example, if a
function takes in an id for a document in the 'users' table, take in
Id<'users'>rather thanstring. - Always use
as constfor string literals in discriminated union types. - When using the
Arraytype, make sure to always define your arrays asconst array: Array<T> = [...]; - When using the
Recordtype, make sure to always define your records asconst record: Record<KeyType, ValueType> = {...}; - Always add
@types/nodeto yourpackage.jsonwhen using any Node.js built-in modules.
Full text search guidelines
- A query for "10 messages in channel '#general' that best match the query 'hello hi' in their body" would look like:
const messages = await ctx.db .query("messages") .withSearchIndex("search_body", (q) => q.search("body", "hello hi").eq("channel", "#general"), ) .take(10);
Query guidelines
- Do NOT use
filterin queries. Instead, define an index in the schema and usewithIndexinstead. - Convex queries do NOT support
.delete(). Instead,.collect()the results, iterate over them, and callctx.db.delete(row._id)on each result. - Use
.unique()to get a single document from a query. This method will throw an error if there are multiple documents that match the query. - When using async iteration, don't use
.collect()or.take(n)on the result of a query. Instead, use thefor await (const row of query)syntax.
Ordering
- By default Convex always returns documents in ascending
_creationTimeorder. - You can use
.order('asc')or.order('desc')to pick whether a query is in ascending or descending order. If the order isn't specified, it defaults to ascending. - Document queries that use indexes will be ordered based on the columns in the index and can avoid slow table scans.
Mutation guidelines
- Use
ctx.db.replaceto fully replace an existing document. This method will throw an error if the document does not exist. Syntax:await ctx.db.replace('tasks', taskId, { name: 'Buy milk', completed: false }) - Use
ctx.db.patchto shallow merge updates into an existing document. This method will throw an error if the document does not exist. Syntax:await ctx.db.patch('tasks', taskId, { completed: true })
Action guidelines
- Always add
"use node";to the top of files containing actions that use Node.js built-in modules. - Never use
ctx.dbinside of an action. Actions don't have access to the database. - Below is an example of the syntax for an action:
import {action} from "./_generated/server";
export const exampleAction = action({
args: {},
returns: v.null(),
handler: async (ctx, args) => {
console.log("This action does not return anything");
return null;
},
});
Scheduling guidelines
Cron guidelines
- Only use the
crons.intervalorcrons.cronmethods to schedule cron jobs. Do NOT use thecrons.hourly,crons.daily, orcrons.weeklyhelpers. - Both cron methods take in a FunctionReference. Do NOT try to pass the function directly into one of these methods.
- Define crons by declaring the top-level
cronsobject, calling some methods on it, and then exporting it as default. For example,
import {cronJobs} from "convex/server";
import {internal} from "./_generated/api";
import {internalAction} from "./_generated/server";
const empty = internalAction({
args: {},
returns: v.null(),
handler: async (ctx, args) => {
console.log("empty");
},
});
const crons = cronJobs();
// Run `internal.crons.empty` every two hours.
crons.interval("delete inactive users", {hours: 2}, internal.crons.empty, {});
export default crons;
- You can register Convex functions within
crons.tsjust like any other file. - If a cron calls an internal function, always import the
internalobject from '_ generated/api', even if the internal function is registered in the same file.
File storage guidelines
-
Convex includes file storage for large files like images, videos, and PDFs.
-
The
ctx.storage.getUrl()method returns a signed URL for a given file. It returnsnullif the file doesn't exist. -
Do NOT use the deprecated
ctx.storage.getMetadatacall for loading a file's metadata.Instead, query the `_storage` system table. For example, you can use `ctx.db.system.get` to get an `Id<"_storage">`.
import { query } from "./_generated/server";
import { Id } from "./_generated/dataModel";
type FileMetadata = {
_id: Id<"_storage">;
_creationTime: number;
contentType?: string;
sha256: string;
size: number;
}
export const exampleQuery = query({
args: { fileId: v.id("_storage") },
returns: v.null(),
handler: async (ctx, args) => {
const metadata: FileMetadata | null = await ctx.db.system.get("_storage", args.fileId);
console.log(metadata);
return null;
},
});
- Convex storage stores items as
Blobobjects. You must convert all items to/from aBlobwhen using Convex storage.