Skip to content

Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure

TypeScript

fix(mcp-host): make csv local processing guidance explicit

Alfredo Lopez09a55fb82ab8
diff --git a/mcp-host/package-lock.json b/mcp-host/package-lock.json
index abfaeaddd..72e547824 100644
--- a/mcp-host/package-lock.json
+++ b/mcp-host/package-lock.json
@@ -1,12 +1,12 @@
{
"name": "clerum-mcp-host",
- "version": "0.1.496",
+ "version": "0.1.497",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "clerum-mcp-host",
- "version": "0.1.496",
+ "version": "0.1.497",
"license": "SEE LICENSE IN LICENSE",
"dependencies": {
"@anthropic-ai/sdk": "^0.32.0",
diff --git a/mcp-host/package.json b/mcp-host/package.json
index 4c3a21aba..986221fc3 100644
--- a/mcp-host/package.json
+++ b/mcp-host/package.json
@@ -1,6 +1,6 @@
{
"name": "clerum-mcp-host",
- "version": "0.1.496",
+ "version": "0.1.497",
"description": "MCP Host - reads Host CRD and provides LLM access via OpenAI/Claude",
"main": "dist/main.js",
"engines": {
diff --git a/mcp-host/src/core/tools/__tests__/shell.test.ts b/mcp-host/src/core/tools/__tests__/shell.test.ts
index e147698bb..4e089e12d 100644
--- a/mcp-host/src/core/tools/__tests__/shell.test.ts
+++ b/mcp-host/src/core/tools/__tests__/shell.test.ts
@@ -84,6 +84,8 @@ describe('ShellTool', () => {
const resolverBase = JSON.parse(
description.match(/createRequire\(("(?:[^"\\]|\\.)*")\)/)![1]
) as string
+ expect(description).toContain('load fast-csv through that resolver and use parseStream')
+ expect(description).toContain('exceljs.csv is not available')
await writeFile(
join(workspacePath, 'input.csv'),
'id,"label,name",notes\r\n' +
diff --git a/mcp-host/src/core/tools/shell.ts b/mcp-host/src/core/tools/shell.ts
index c1010eb7a..dc861f55f 100644
--- a/mcp-host/src/core/tools/shell.ts
+++ b/mcp-host/src/core/tools/shell.ts
@@ -104,7 +104,7 @@ export class ShellTool implements Tool {
'Supports full shell syntax (pipes, redirects, &&, etc.). ' +
'Commands run with a timeout and restricted environment. ' +
`Node.js executable: ${JSON.stringify(process.execPath)}. ` +
- `Resolve installed Host libraries with require('node:module').createRequire(${JSON.stringify(require.resolve('exceljs'))}); the existing exceljs dependency includes fast-csv for streaming CSV parsing. ` +
+ `Resolve installed Host libraries with require('node:module').createRequire(${JSON.stringify(require.resolve('exceljs'))}); load fast-csv through that resolver and use parseStream for streaming CSV parsing (exceljs.csv is not available). ` +
'Verify any other executable or library before using it. ' +
'This tool requires approval before execution.'
)
diff --git a/mcp-host/src/internalTools/gfsReadTypes.ts b/mcp-host/src/internalTools/gfsReadTypes.ts
index 61cfeb78f..f2167bcd9 100644
--- a/mcp-host/src/internalTools/gfsReadTypes.ts
+++ b/mcp-host/src/internalTools/gfsReadTypes.ts
@@ -4,6 +4,7 @@ import type { GfsImageSource, MemoryReservation, VisualInputBudget } from '../vi
export const GFS_LOCAL_PROCESSING_GUIDANCE =
'Use the Node executable and installed-library resolver advertised by shell_exec; verify any other executable or library instead of assuming Python or additional packages are installed. ' +
'Prefer an installed streaming parser for the actual file format and keep memory and output bounded. ' +
+ 'For CSV, load fast-csv through that resolver and use its parseStream API; this installation does not expose exceljs.csv. ' +
'Count logical records and parse fields according to that format: CSV can contain quoted delimiters, escaped quotes and embedded newlines, so splitting on commas or counting physical lines is not a CSV parser. ' +
'Have the script compute and label every reported quantity, including field counts; do not estimate or recount an array in your reply. ' +
'Return all requested metadata names within the output budget. A brief summary must still answer every requested part; do not defer an already requested list to another message. ' +