Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
docs(mcp-host): correct attachment limits and body budget comments
Alfredo Lopez120cb362bb9e
diff --git a/mcp-host/README.md b/mcp-host/README.mdindex 8e3e74eb8..7bc8a86e7 100644--- a/mcp-host/README.md+++ b/mcp-host/README.md@@ -55,7 +55,8 @@ MCP Host is a Kubernetes-native service that reads Host CRD configuration and pr| `CLERUM_ENABLE_RESPONSE_ATTACHMENTS` | Enables attachment passthrough in runtime responses when tools return supported media. | `CLERUM_ENABLE_RESPONSE_ATTACHMENTS=true` || `CLERUM_ATTACHMENT_MAX_COUNT` | Maximum number of attachments returned per response. | `CLERUM_ATTACHMENT_MAX_COUNT=3` || `CLERUM_ATTACHMENT_MAX_BYTES` | Maximum decoded size per attachment in bytes (oversized payloads are dropped). | `CLERUM_ATTACHMENT_MAX_BYTES=52428800` |-| `CLERUM_ATTACHMENT_FILE_MAX_BYTES` | Maximum decoded size in bytes of one incoming `kind:'file'` attachment; a larger file rejects the message with `FILE_ATTACHMENT_TOO_LARGE`. An invalid value stops the Host at startup. | `CLERUM_ATTACHMENT_FILE_MAX_BYTES=3145728` |+| `CLERUM_ATTACHMENT_FILE_MAX_BYTES` | Maximum decoded size in bytes of one incoming `kind:'file'` attachment; one larger file rejects the message with `FILE_ATTACHMENT_TOO_LARGE`. File base64 also counts against the 6 MiB non-image body budget, so two files at the 3 MiB default get HTTP 413 before admission. An invalid value stops the Host at startup. | `CLERUM_ATTACHMENT_FILE_MAX_BYTES=3145728` |+| `CLERUM_FILE_REFERENCE_MAX_COUNT` | Maximum number of `fileReferences` (structured Global Files selections, #666) one message may carry; more rejects the message with `FILE_REFERENCE_INVALID`. An invalid value stops the Host at startup. | `CLERUM_FILE_REFERENCE_MAX_COUNT=10` || `CLERUM_ATTACHMENT_TEXT_READ_MAX_BYTES` | Maximum bytes one `clerum__attachment_read` call returns; longer text is returned with `truncated:true` and read on with `offset`. An invalid value stops the Host at startup. | `CLERUM_ATTACHMENT_TEXT_READ_MAX_BYTES=262144` || `CLERUM_MEMORY_ENABLED` | Enables workspace memory persistence features. | `CLERUM_MEMORY_ENABLED=true` || `CLERUM_MEMORY_WORKSPACE_PATH` | Sets workspace directory path used when memory mode is enabled. | `CLERUM_MEMORY_WORKSPACE_PATH=/workspace` |diff --git a/mcp-host/package-lock.json b/mcp-host/package-lock.jsonindex 0854ac125..144427752 100644--- a/mcp-host/package-lock.json+++ b/mcp-host/package-lock.json@@ -1,12 +1,12 @@{"name": "clerum-mcp-host",- "version": "0.1.451",+ "version": "0.1.452","lockfileVersion": 3,"requires": true,"packages": {"": {"name": "clerum-mcp-host",- "version": "0.1.451",+ "version": "0.1.452","license": "SEE LICENSE IN LICENSE","dependencies": {"@anthropic-ai/sdk": "^0.32.0",diff --git a/mcp-host/package.json b/mcp-host/package.jsonindex 133f9ab58..6a7083d84 100644--- a/mcp-host/package.json+++ b/mcp-host/package.json@@ -1,6 +1,6 @@{"name": "clerum-mcp-host",- "version": "0.1.451",+ "version": "0.1.452","description": "MCP Host - reads Host CRD and provides LLM access via OpenAI/Claude","main": "dist/main.js","engines": {diff --git a/mcp-host/src/server.ts b/mcp-host/src/server.tsindex 9d45efb78..9da365783 100644--- a/mcp-host/src/server.ts+++ b/mcp-host/src/server.ts@@ -129,18 +129,23 @@ export type {} from './server/types'/**- * Body budgets for chat payloads carrying base64 image attachments.+ * Body budgets for chat payloads carrying base64 attachments.*- * Intentional mirror of `rpc-proxy/src/app.ts`: rpc-proxy has no dependency on- * any @clerum package, so the two services cannot share one source without a- * new package. Both copies must change together.+ * Intentional mirror of `rpc-proxy/src/middleware/chatJsonBody.ts`: rpc-proxy+ * has no dependency on any @clerum package, so the two services cannot share+ * one source without a new package. Both copies must change together.** - MAX_CHAT_BODY_BYTES stays 24MiB so one exceptional 16MiB 2048 PNG* (~21.3MiB base64) plus the 1MiB non-image share still fits. Every other- * route keeps the 6MiB default.+ * route keeps the 10mb ordinary JSON cap (see the parsers below).* - MAX_NON_IMAGE_BODY_BYTES bounds that same body MINUS credited image* base64 (16MiB per image, at most 20 images / 16MiB total). Usual product* target remains 5 / 9 / 14 MiB at 2048 px.+ * - `kind:'file'` attachments are never credited: their base64 (4/3 of the+ * decoded size) counts against MAX_NON_IMAGE_BODY_BYTES with the rest of+ * the body. One file at the 3MiB CLERUM_ATTACHMENT_FILE_MAX_BYTES default+ * (4MiB base64) fits; two do not, and the message gets this 413 before+ * admission can answer FILE_ATTACHMENT_TOO_LARGE for either file.*/const MAX_CHAT_BODY_BYTES = 24 * 1024 * 1024const MAX_NON_IMAGE_BODY_BYTES = 6 * 1024 * 1024@@ -307,8 +312,10 @@ export class RPCServer {next()})- // Image attachments are sent as base64 in /v1/runtime/messages.- // That route alone gets the 24 MiB ceiling. Every other Host route uses+ // Image and file attachments are sent as base64 in /v1/runtime/messages.+ // That route alone gets the 24 MiB ceiling; within it, only qualifying+ // images are credited, and `kind:'file'` base64 is charged to the 6 MiB+ // non-image budget like any other field. Every other Host route uses// the same 10mb ordinary JSON cap as rpc-proxy `jsonBody`, so a future// non-chat control body cannot 413 here and pass the proxy.const jsonParser = express.json({ limit: '10mb' })