secret scanner

README.md

@@ -291,6 +291,19 @@ Write me a poem
 
 ---
 
+### Pluggable Secret Scanning
+
+Scan your chats for secrets using [secret scanning](/genaiscript/reference/scripts/secret-scanning).
+
+```json
+{
+    "secretPatterns": {
+        ...,
+        "OpenAI API Key": "sk-[A-Za-z0-9]{32,48}"
+    }
+}
+```
+
 ### ⚙ Automate with CLI or API
 
 Automate using the [CLI](https://microsoft.github.io/genaiscript/reference/cli) or [API](https://microsoft.github.io/genaiscript/reference/cli/api).

docs/public/schemas/config.json

@@ -40,7 +40,7 @@
         "modelAliases": {
             "type": "object",
             "patternProperties": {
-                "^[a-zA-Z0-9_:]+$": {
+                "^[a-zA-Z0-9_]+$": {
                     "oneOf": [
                         {
                             "type": "string",
@@ -64,7 +64,18 @@
                 }
             },
             "additionalProperties": true,
-            "description": "Aliases for model identifiers (provider:model:tag)"
+            "description": "Aliases for model identifiers (name)"
+        },
+        "secretPatterns": {
+            "type": "object",
+            "patternProperties": {
+                "^[a-zA-Z0-9_:\\-\\. ]+$": {
+                    "type": ["string", "null"],
+                    "description": "Secret regex"
+                }
+            },
+            "additionalProperties": true,
+            "description": "Secret scanners to use for scanning chat messages"
         }
     }
 }

docs/src/content/docs/index.mdx

@@ -360,7 +360,7 @@ to validate [content safety](/genaiscript/reference/scripts/content-safety).
 
 ```js wrap
 script({ ...,
-    system: ["system.safety_harmful_content", ...],
+    systemSafety: "default",
     contentSafety: "azure" // use azure content safety
 })
 
@@ -486,6 +486,21 @@ importTemplate("poem.prompty", { something: "code " })
 
 </Card>
 
+<Card title="Pluggable Secret Scanning" icon="seti:license">
+
+Scan your chats for secrets using [secret scanning](/genaiscript/reference/scripts/secret-scanning).
+
+```json title="genaiscript.config.json"
+{
+    "secretPatterns": {
+        ...,
+        "OpenAI API Key": "sk-[A-Za-z0-9]{32,48}"
+    }
+}
+```
+
+</Card>
+
 <Card title="Automate with CLI" icon="github">
 
 Automate using the [CLI](/genaiscript/reference/cli),

docs/src/content/docs/reference/cli/commands.md

@@ -452,6 +452,7 @@ Commands:
   jsonl2json                   Converts JSONL files to a JSON file
   prompty [options] <file...>  Converts .prompty files to genaiscript
   jinja2 [options] <file>      Renders Jinj2 or prompty template
+  secrets <file...>            Applies secret scanning and redaction to files
 ```
 
 ### `parse data`
@@ -578,6 +579,20 @@ Options:
   -h, --help             display help for command
 ```
 
+### `parse secrets`
+
+```
+Usage: genaiscript parse secrets [options] <file...>
+
+Applies secret scanning and redaction to files
+
+Arguments:
+  file        input files
+
+Options:
+  -h, --help  display help for command
+```
+
 ## `info`
 
 ```

docs/src/content/docs/reference/scripts/content-safety.mdx

@@ -16,6 +16,14 @@ The following safety prompts are included by default when running a prompt, unle
 - [system.safety_jailbreak](/genaiscript/reference/scripts/system#systemsafety_jailbreak), safety script to ignore prompting instructions in code sections, which are created by the `def` function.
 - [system.safety_protected_material](/genaiscript/reference/scripts/system#systemsafety_protected_material) safety prompt against Protected material. See https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/safety-system-message-templates
 
+You can ensure those safety are always used by setting the `systemSafety` option to `default`.
+
+```js
+script({
+    systemSafety: "default",
+})
+```
+
 Other system scripts can be added to the prompt by using the `system` option.
 
 - [system.safety_ungrounded_content_summarization](/genaiscript/reference/scripts/system#systemsafety_ungrounded_content_summarization) safety prompt against ungrounded content in summarization

docs/src/content/docs/reference/scripts/images.md

@@ -30,7 +30,7 @@ Local files are loaded and encoded as a data uri.
 
 ## Buffer, Blob, ReadableStream
 
-The `defImages` function also supports [Buffer](https://nodejs.org/api/buffer.html), 
+The `defImages` function also supports [Buffer](https://nodejs.org/api/buffer.html),
 [Blob](https://developer.mozilla.org/en-US/docs/Web/API/Blob), [ReadableStream](https://nodejs.org/api/stream.html).
 
 This example takes a screenshot of bing.com and adds it to the images.
@@ -108,3 +108,14 @@ defImages(img, { maxWidth: 800 })
 // and / or
 defImages(img, { maxHeight: 800 })
 ```
+
+## Tiling
+
+When you specify the `tiled: true` option, all the images will
+be tiled in a single image, after all the transformations are applied.
+
+The resulting image will be further resized to fit into the maximum image size constraints.
+
+```js "tiled: true"
+defImages(env.files, { details: "low", tiled: true })
+```

docs/src/content/docs/reference/scripts/secret-scanning.mdx

@@ -0,0 +1,77 @@
+---
+title: Secret Scanning
+sidebar:
+    order: 10
+---
+
+One should not have secrets lying around in their codebase, but sometimes it happens.
+To help you avoid this, we have a secret scanning feature that will scan your codebase for secrets
+and warn you if any are found.
+
+:::note
+
+The secret scanning feature is by no means exhaustive and should not be relied upon as the sole
+method of securing your codebase. It is a best-effort feature that will help you avoid common mistakes.
+
+:::
+
+## Supported patterns
+
+By default set of secret patterns
+are defined at https://github.com/microsoft/genaiscript/tree/main/packages/core/src/config.json.
+
+:::cautio
+
+\is is not a complete list by design, and needs to be updated to match your needs.
+
+:::
+
+You can find examples of patterns at https://github.com/mazen160/secrets-patterns-db/.
+
+## Scanning messages
+
+By default, all messages sent to LLMs are scanned and redacted if they contain secrets.
+
+You can disable secret scanning alltogher by setting the `secretScanning` option to `false` in your script.
+
+```js
+script({
+    secretScanning: false,
+})
+```
+
+## Configuring patterns
+
+If you have a specific pattern that you want to scan for, you can configure it in your
+[configuration file](/genaiscript/reference/scripts/configuration-files).
+
+```json title="genaiscript.config.json"
+{
+    "secretPatterns": {
+        ...,
+        "my secret pattern": "my-secret-pattern-regex"
+    }
+}
+```
+
+- do not use `^` or `$` in your regex pattern
+
+### Disabling patterns
+
+Set the pattern key to `null` or `false` to disable it.
+
+```json title="genaiscript.config.json"
+{
+    "secretPatterns": {
+        "OpenAI API Key": null
+    }
+}
+```
+
+## CLI
+
+You can test your patterns against files using the CLI.
+
+```sh
+npx --yes genaiscript parse secrets *
+```

docs/src/content/docs/reference/scripts/system.mdx

@@ -3463,7 +3463,8 @@ system({
     title: "Tools support",
 })
 
-$`Use tools if possible. 
+$`## Tools
+Use tools if possible. 
 - **Do NOT invent function names**. 
 - **Do NOT use function names starting with 'functions.'.
 - **Do NOT respond with multi_tool_use**.`

packages/cli/src/cli.ts

@@ -19,6 +19,7 @@ import {
     parseHTMLToText,
     parseJinja2,
     parsePDF,
+    parseSecrets,
     parseTokens,
     prompty2genaiscript,
 } from "./parse" // Parsing functions
@@ -479,6 +480,11 @@ export async function cli() {
             "variables, as name=value passed to the template"
         )
         .action(parseJinja2)
+    parser
+        .command("secrets")
+        .description("Applies secret scanning and redaction to files")
+        .argument("<file...>", "input files")
+        .action(parseSecrets)
 
     // Define 'info' command group for utility information tasks
     const info = program.command("info").description("Utility tasks")

packages/cli/src/parse.ts

@@ -25,6 +25,7 @@ import { splitMarkdown } from "../../core/src/frontmatter"
 import { parseOptionsVars } from "./vars"
 import { dataTryParse } from "../../core/src/data"
 import { resolveFileContent } from "../../core/src/file"
+import { redactSecrets } from "../../core/src/secretscanner"
 
 /**
  * This module provides various parsing utilities for different file types such
@@ -215,3 +216,20 @@ export async function prompty2genaiscript(
         await writeText(gf, script)
     }
 }
+
+export async function parseSecrets(files: string[]) {
+    const fs = await expandFiles(files)
+    let n = 0
+    for (const f of fs) {
+        const content = await readText(f)
+        const { found } = redactSecrets(content)
+        const entries = Object.entries(found)
+        if (entries.length) {
+            n++
+            console.log(
+                `${f}: ${entries.map(([k, v]) => `${k} (${v})`).join(", ")}`
+            )
+        }
+    }
+    if (n > 0) console.warn(`found secrets in ${n} of ${fs.length} files`)
+}

packages/core/src/config.json

@@ -0,0 +1,6 @@
+{
+    "$schema": "../../../docs/public/schemas/config.json",
+    "secretPatterns": {
+        "OpenAI API Key": "sk-[A-Za-z0-9]{32,48}"
+    }
+}

packages/core/src/config.ts

@@ -17,6 +17,7 @@ import { resolveLanguageModel } from "./lm"
 import { deleteEmptyValues } from "./cleaners"
 import { errorMessage } from "./error"
 import schema from "../../../docs/public/schemas/config.json"
+import defaultConfig from "./config.json"
 
 export async function resolveGlobalConfiguration(
     dotEnvPath?: string
@@ -27,13 +28,14 @@ export async function resolveGlobalConfiguration(
     const exts = ["yml", "yaml", "json"]
 
     // import and merge global local files
-    let config: HostConfiguration = {}
+    let config: HostConfiguration = structuredClone(defaultConfig)
+    delete (config as any)["$schema"]
     for (const dir of dirs) {
         for (const ext of exts) {
             const filename = resolve(`${dir}/${TOOL_ID}.config.${ext}`)
             if (existsSync(filename)) {
                 const fileContent = readFileSync(filename, "utf8")
-                const parsed =
+                const parsed: HostConfiguration =
                     ext === "yml" || ext === "yaml"
                         ? YAMLTryParse(fileContent)
                         : JSON5TryParse(fileContent)
@@ -63,6 +65,10 @@ export async function resolveGlobalConfiguration(
                         config?.modelEncodings || {},
                         parsed?.modelEncodings || {}
                     ),
+                    secretScanners: structuralMerge(
+                        config?.secretPatterns || {},
+                        parsed?.secretPatterns || {}
+                    ),
                 })
             }
         }

packages/core/src/expander.ts

@@ -17,10 +17,7 @@ import {
 import { createPromptContext } from "./promptcontext"
 import { evalPrompt } from "./evalprompt"
 import { renderAICI } from "./aici"
-import {
-    addToolDefinitionsMessage,
-    appendSystemMessage,
-} from "./chat"
+import { addToolDefinitionsMessage, appendSystemMessage } from "./chat"
 import { importPrompt } from "./importprompt"
 import { parseModelIdentifier } from "./models"
 import { runtimeHost } from "./host"
@@ -355,8 +352,7 @@ export async function expandTemplate(
     }
 
     const { responseType, responseSchema } = finalizeMessages(messages, {
-        responseType: template.responseType,
-        responseSchema: template.responseSchema,
+        ...template,
         fileOutputs,
         trace,
     })

packages/core/src/genaisrc/system.tools.genai.js

@@ -2,7 +2,8 @@ system({
     title: "Tools support",
 })
 
-$`Use tools if possible. 
+$`## Tools
+Use tools if possible. 
 - **Do NOT invent function names**. 
 - **Do NOT use function names starting with 'functions.'.
 - **Do NOT respond with multi_tool_use**.`

packages/core/src/hostconfiguration.ts

@@ -23,4 +23,9 @@ export interface HostConfiguration {
      * Model identifier to encoding mapping
      */
     modelEncodings?: Record<string, string>
+
+    /**
+     * A map of secret name and their respective regex pattern
+     */
+    secretPatterns?: Record<string, string>
 }