Better memory management

Added official file support
2026-08-24 14:42:10 -04:00 · 2026-08-17 15:50:48 -04:00 · 2026-08-17 15:16:32 -04:00 · 2026-08-16 15:40:50 -04:00
6 changed files with 601 additions and 246 deletions
--- a/package-lock.json
+++ b/package-lock.json
@@ -1,12 +1,12 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "1.2.6",
+	"version": "1.5.0",
 	"lockfileVersion": 3,
 	"requires": true,
 	"packages": {
 		"": {
 			"name": "@ztimson/ai-utils",
-			"version": "1.2.6",
+			"version": "1.5.0",
 			"license": "MIT",
 			"dependencies": {
 				"@anthropic-ai/sdk": "^0.102.0",
@@ -16,6 +16,7 @@
 				"@ztimson/utils": "^0.29.4",
 				"cheerio": "^1.2.0",
 				"openai": "^6.42.0",
+				"pdf-parse": "^2.4.5",
 				"tesseract.js": "^7.0.0"
 			},
 			"devDependencies": {
@@ -56,39 +57,12 @@
 				"node": ">=6.9.0"
 			}
 		},
-		"node_modules/@emnapi/core": {
-			"version": "2.0.0-alpha.3",
-			"resolved": "https://registry.npmjs.org/@emnapi/core/-/core-2.0.0-alpha.3.tgz",
-			"integrity": "sha512-AZypUeJ/yByuxyS7BlSNRDOMLMlROYtjYdIAuBmJssVz1UJDSeYxLrdizhXCFYhedC5bqd/ASy8EuNXbVVXp9g==",
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"peer": true,
-			"dependencies": {
-				"@emnapi/wasi-threads": "2.0.1",
-				"tslib": "^2.4.0"
-			}
-		},
 		"node_modules/@emnapi/runtime": {
-			"version": "2.0.0-alpha.3",
-			"resolved": "https://registry.npmjs.org/@emnapi/runtime/-/runtime-2.0.0-alpha.3.tgz",
-			"integrity": "sha512-hFPAhMUjJD9BSyCANEISPOogeXC9Zo9ZQl7L6vKnaVsMkCtzznaW/naYypeyl0Gv5rYfWYsZbpixTMpjDJzQeA==",
-			"dev": true,
+			"version": "1.11.3",
+			"resolved": "https://registry.npmjs.org/@emnapi/runtime/-/runtime-1.11.3.tgz",
+			"integrity": "sha512-Xz4Tpyki7XyrpbUK1jR1AhdAdaXyhhY4lZ3neLodmhpuWfy2PAQN5B46sAiU4liOXGLkHypn/qU+jvfWSCYYLA==",
 			"license": "MIT",
 			"optional": true,
-			"peer": true,
-			"dependencies": {
-				"tslib": "^2.4.0"
-			}
-		},
-		"node_modules/@emnapi/wasi-threads": {
-			"version": "2.0.1",
-			"resolved": "https://registry.npmjs.org/@emnapi/wasi-threads/-/wasi-threads-2.0.1.tgz",
-			"integrity": "sha512-9DsSk+o5NBX0CCJT8s0EROGSGxjR/tKu6aBTaVyq+SjAEQH4XcdcRxPBRzsBLizTTJ49MJjF+jgu3qnO9GLQcQ==",
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"peer": true,
 			"dependencies": {
 				"tslib": "^2.4.0"
 			}
@@ -577,16 +551,6 @@
 				"url": "https://opencollective.com/libvips"
 			}
 		},
-		"node_modules/@img/sharp-wasm32/node_modules/@emnapi/runtime": {
-			"version": "1.11.3",
-			"resolved": "https://registry.npmjs.org/@emnapi/runtime/-/runtime-1.11.3.tgz",
-			"integrity": "sha512-Xz4Tpyki7XyrpbUK1jR1AhdAdaXyhhY4lZ3neLodmhpuWfy2PAQN5B46sAiU4liOXGLkHypn/qU+jvfWSCYYLA==",
-			"license": "MIT",
-			"optional": true,
-			"dependencies": {
-				"tslib": "^2.4.0"
-			}
-		},
 		"node_modules/@img/sharp-win32-arm64": {
 			"version": "0.34.5",
 			"resolved": "https://registry.npmjs.org/@img/sharp-win32-arm64/-/sharp-win32-arm64-0.34.5.tgz",
@@ -694,32 +658,209 @@
 				"@jridgewell/sourcemap-codec": "^1.4.14"
 			}
 		},
-		"node_modules/@napi-rs/wasm-runtime": {
-			"version": "1.2.0",
-			"resolved": "https://registry.npmjs.org/@napi-rs/wasm-runtime/-/wasm-runtime-1.2.0.tgz",
-			"integrity": "sha512-kDoONqMa+VnZ4vvvu/ZUurpJ4gkZU57e7g69qpNgWhYcZFPUHZM2CEMKm+cG6ufDVALbjMvfmMjFVqaK7uEMnA==",
-			"dev": true,
+		"node_modules/@napi-rs/canvas": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas/-/canvas-0.1.80.tgz",
+			"integrity": "sha512-DxuT1ClnIPts1kQx8FBmkk4BQDTfI5kIzywAaMjQSXfNnra5UFU9PwurXrl+Je3bJ6BGsp/zmshVVFbCmyI+ww==",
+			"license": "MIT",
+			"workspaces": [
+				"e2e/*"
+			],
+			"engines": {
+				"node": ">= 10"
+			},
+			"optionalDependencies": {
+				"@napi-rs/canvas-android-arm64": "0.1.80",
+				"@napi-rs/canvas-darwin-arm64": "0.1.80",
+				"@napi-rs/canvas-darwin-x64": "0.1.80",
+				"@napi-rs/canvas-linux-arm-gnueabihf": "0.1.80",
+				"@napi-rs/canvas-linux-arm64-gnu": "0.1.80",
+				"@napi-rs/canvas-linux-arm64-musl": "0.1.80",
+				"@napi-rs/canvas-linux-riscv64-gnu": "0.1.80",
+				"@napi-rs/canvas-linux-x64-gnu": "0.1.80",
+				"@napi-rs/canvas-linux-x64-musl": "0.1.80",
+				"@napi-rs/canvas-win32-x64-msvc": "0.1.80"
+			}
+		},
+		"node_modules/@napi-rs/canvas-android-arm64": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-android-arm64/-/canvas-android-arm64-0.1.80.tgz",
+			"integrity": "sha512-sk7xhN/MoXeuExlggf91pNziBxLPVUqF2CAVnB57KLG/pz7+U5TKG8eXdc3pm0d7Od0WreB6ZKLj37sX9muGOQ==",
+			"cpu": [
+				"arm64"
+			],
 			"license": "MIT",
 			"optional": true,
-			"dependencies": {
-				"@tybys/wasm-util": "^0.10.3"
-			},
+			"os": [
+				"android"
+			],
 			"engines": {
-				"node": "^20.19.0 || ^22.13.0 || >=23.5.0"
+				"node": ">= 10"
+			}
 		},
-			"funding": {
-				"type": "github",
-				"url": "https://github.com/sponsors/Brooooooklyn"
+		"node_modules/@napi-rs/canvas-darwin-arm64": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-darwin-arm64/-/canvas-darwin-arm64-0.1.80.tgz",
+			"integrity": "sha512-O64APRTXRUiAz0P8gErkfEr3lipLJgM6pjATwavZ22ebhjYl/SUbpgM0xcWPQBNMP1n29afAC/Us5PX1vg+JNQ==",
+			"cpu": [
+				"arm64"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"darwin"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
 		},
-			"peerDependencies": {
-				"@emnapi/core": "^2.0.0-alpha.3",
-				"@emnapi/runtime": "^2.0.0-alpha.3"
+		"node_modules/@napi-rs/canvas-darwin-x64": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-darwin-x64/-/canvas-darwin-x64-0.1.80.tgz",
+			"integrity": "sha512-FqqSU7qFce0Cp3pwnTjVkKjjOtxMqRe6lmINxpIZYaZNnVI0H5FtsaraZJ36SiTHNjZlUB69/HhxNDT1Aaa9vA==",
+			"cpu": [
+				"x64"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"darwin"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-linux-arm-gnueabihf": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-linux-arm-gnueabihf/-/canvas-linux-arm-gnueabihf-0.1.80.tgz",
+			"integrity": "sha512-eyWz0ddBDQc7/JbAtY4OtZ5SpK8tR4JsCYEZjCE3dI8pqoWUC8oMwYSBGCYfsx2w47cQgQCgMVRVTFiiO38hHQ==",
+			"cpu": [
+				"arm"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"linux"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-linux-arm64-gnu": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-linux-arm64-gnu/-/canvas-linux-arm64-gnu-0.1.80.tgz",
+			"integrity": "sha512-qwA63t8A86bnxhuA/GwOkK3jvb+XTQaTiVML0vAWoHyoZYTjNs7BzoOONDgTnNtr8/yHrq64XXzUoLqDzU+Uuw==",
+			"cpu": [
+				"arm64"
+			],
+			"libc": [
+				"glibc"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"linux"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-linux-arm64-musl": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-linux-arm64-musl/-/canvas-linux-arm64-musl-0.1.80.tgz",
+			"integrity": "sha512-1XbCOz/ymhj24lFaIXtWnwv/6eFHXDrjP0jYkc6iHQ9q8oXKzUX1Lc6bu+wuGiLhGh2GS/2JlfORC5ZcXimRcg==",
+			"cpu": [
+				"arm64"
+			],
+			"libc": [
+				"musl"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"linux"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-linux-riscv64-gnu": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-linux-riscv64-gnu/-/canvas-linux-riscv64-gnu-0.1.80.tgz",
+			"integrity": "sha512-XTzR125w5ZMs0lJcxRlS1K3P5RaZ9RmUsPtd1uGt+EfDyYMu4c6SEROYsxyatbbu/2+lPe7MPHOO/0a0x7L/gw==",
+			"cpu": [
+				"riscv64"
+			],
+			"libc": [
+				"glibc"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"linux"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-linux-x64-gnu": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-linux-x64-gnu/-/canvas-linux-x64-gnu-0.1.80.tgz",
+			"integrity": "sha512-BeXAmhKg1kX3UCrJsYbdQd3hIMDH/K6HnP/pG2LuITaXhXBiNdh//TVVVVCBbJzVQaV5gK/4ZOCMrQW9mvuTqA==",
+			"cpu": [
+				"x64"
+			],
+			"libc": [
+				"glibc"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"linux"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-linux-x64-musl": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-linux-x64-musl/-/canvas-linux-x64-musl-0.1.80.tgz",
+			"integrity": "sha512-x0XvZWdHbkgdgucJsRxprX/4o4sEed7qo9rCQA9ugiS9qE2QvP0RIiEugtZhfLH3cyI+jIRFJHV4Fuz+1BHHMg==",
+			"cpu": [
+				"x64"
+			],
+			"libc": [
+				"musl"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"linux"
+			],
+			"engines": {
+				"node": ">= 10"
+			}
+		},
+		"node_modules/@napi-rs/canvas-win32-x64-msvc": {
+			"version": "0.1.80",
+			"resolved": "https://registry.npmjs.org/@napi-rs/canvas-win32-x64-msvc/-/canvas-win32-x64-msvc-0.1.80.tgz",
+			"integrity": "sha512-Z8jPsM6df5V8B1HrCHB05+bDiCxjE9QA//3YrkKIdVDEwn5RKaqOxCJDRJkl48cJbylcrJbW4HxZbTte8juuPg==",
+			"cpu": [
+				"x64"
+			],
+			"license": "MIT",
+			"optional": true,
+			"os": [
+				"win32"
+			],
+			"engines": {
+				"node": ">= 10"
 			}
 		},
 		"node_modules/@oxc-project/types": {
-			"version": "0.139.0",
-			"resolved": "https://registry.npmjs.org/@oxc-project/types/-/types-0.139.0.tgz",
-			"integrity": "sha512-r9gHphtCs+1M7J0pw6Sn/hh/Wpa/iQrOOkrNAlVLF/gHq+/CJmHIWKKUUhdWjcD6CIa8idarspCsASiXCXvFUw==",
+			"version": "0.144.0",
+			"resolved": "https://registry.npmjs.org/@oxc-project/types/-/types-0.144.0.tgz",
+			"integrity": "sha512-nuhZIOLuI6TFQ32I/WnUx+SCPY7SdSKwgnFHydAuoS1+Z4BRcaP+RRJmGzl9lw+0OFF7UmaESf7KQRXaNLHypg==",
 			"dev": true,
 			"license": "MIT",
 			"funding": {
@@ -784,9 +925,9 @@
 			"license": "BSD-3-Clause"
 		},
 		"node_modules/@rolldown/binding-android-arm64": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-android-arm64/-/binding-android-arm64-1.1.5.tgz",
-			"integrity": "sha512-lZg8fqIv2v7FF237bwMgzGZEJvGL79/s5knJ/i6FmsGF4XXlzccZ4jb+TrFIxtSSxFtIpdsgrPZeMk1I9AFcyQ==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-android-arm64/-/binding-android-arm64-1.2.4.tgz",
+			"integrity": "sha512-jHC2cnyKz5xU2fhECtFl8OZ83cYNt13GZQD+0uMJ/X3o+ijmd56okHhTUwxVSHPx1IRVIJEZ1/1pPzeLCU6XKA==",
 			"cpu": [
 				"arm64"
 			],
@@ -801,9 +942,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-darwin-arm64": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-arm64/-/binding-darwin-arm64-1.1.5.tgz",
-			"integrity": "sha512-51Bnx9pNiMRKSUNtBfySkNJ9vMU9Hh3I1ozDd6gyPPYzaXCfnptUcEZxXGYFn+ul2dtcMUiqGR1Yai2K10uoTw==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-arm64/-/binding-darwin-arm64-1.2.4.tgz",
+			"integrity": "sha512-Dc5mPD8F5F/FS8i01syd7FTF6yB2fVthH/TRkjwJkzUK6EpoxHtqvZQP5Zwq80/5z19TWYHIg1KOHboCgVx/aQ==",
 			"cpu": [
 				"arm64"
 			],
@@ -818,9 +959,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-darwin-x64": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-x64/-/binding-darwin-x64-1.1.5.tgz",
-			"integrity": "sha512-Tm+gbfC0aHu1tBA/JvKQh32S0K6YgCHkiAF4/W6xX0K0RmNuc94VeK419dJoE65R5aRxmo+noZQSWrAMF6yb6g==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-darwin-x64/-/binding-darwin-x64-1.2.4.tgz",
+			"integrity": "sha512-fpDm4oBo6SqLvWUYCmFhdde3U9KH2fRNNMeAnAPAIwxRL345xutL0EtEUcuoxsoazdJGv/MuDBQHlCDrtbvqOg==",
 			"cpu": [
 				"x64"
 			],
@@ -835,9 +976,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-freebsd-x64": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-freebsd-x64/-/binding-freebsd-x64-1.1.5.tgz",
-			"integrity": "sha512-JMzDKCCXq93YccG5gz3hvOs1oXRKAf0XYpfOS88e+wZrC8Iugj6j68867vrYZkvpDDpKn/KoKORThmchMpF6TA==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-freebsd-x64/-/binding-freebsd-x64-1.2.4.tgz",
+			"integrity": "sha512-rSJoreDE/HoIzoaib6MTp5jQtCTdMHKIvItAKT/ImS6Y6Ww76oUaeMyp4Vc/fAgd/ehji068IxetHXAnqUwN9A==",
 			"cpu": [
 				"x64"
 			],
@@ -852,9 +993,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-arm-gnueabihf": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm-gnueabihf/-/binding-linux-arm-gnueabihf-1.1.5.tgz",
-			"integrity": "sha512-uML21j2K5TfPGutKxub+M+nLjZIrWjXQ5Grx4lCe/nimTj9B4L63zHpjXLl4y0L3mcm2htEQIb06oCG/szerNw==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm-gnueabihf/-/binding-linux-arm-gnueabihf-1.2.4.tgz",
+			"integrity": "sha512-/jm8OGHgn7oGaJu3i/qZI9spUGcJ+y/lk43ttQ/iO1tOd9NissG6o97bighBCiL+BKRngmcDuR6ikfwYdJmVuQ==",
 			"cpu": [
 				"arm"
 			],
@@ -869,9 +1010,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-arm64-gnu": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-gnu/-/binding-linux-arm64-gnu-1.1.5.tgz",
-			"integrity": "sha512-navSiuTMogvnQoZoM/v+l3ZWo50/NTwSHSzheABx/RCnmUPaKwq9qSo4Br2OYRs21+Fz8uFqITZM3H4opOB0/Q==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-gnu/-/binding-linux-arm64-gnu-1.2.4.tgz",
+			"integrity": "sha512-tIP06BeD9EqvECBrPZ+sqdPlYrT+aYaAiu1wYziVx5elRK/ftm33JxVDy2bXGbr6J0CrtirCkR87/X5a2euEng==",
 			"cpu": [
 				"arm64"
 			],
@@ -889,9 +1030,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-arm64-musl": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-musl/-/binding-linux-arm64-musl-1.1.5.tgz",
-			"integrity": "sha512-lAryqH7IteztmCXQXk0etKj4wBQ7Gx5S6LjKhsgp9zb8I5bsuvU/2llH1hDQcjsFeqIsovMVN339/8pUDDBXxA==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-arm64-musl/-/binding-linux-arm64-musl-1.2.4.tgz",
+			"integrity": "sha512-Ql1Q0EQqVThvn9VAVlwNzsUvbSFtCMGjLpRRi4pk5i7NZZ4n5ISiLMjHYtus4VQ2PvkSw24zyaCVsiS+sXPj1w==",
 			"cpu": [
 				"arm64"
 			],
@@ -909,9 +1050,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-ppc64-gnu": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-ppc64-gnu/-/binding-linux-ppc64-gnu-1.1.5.tgz",
-			"integrity": "sha512-fsK/sNBnxzBlL4O1JNrZakVQxPspqpED5dLtNsZS9oOKmtSpdNIzxH2kkol5HYTWJN47sE20ztMJPxfZ89qGOg==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-ppc64-gnu/-/binding-linux-ppc64-gnu-1.2.4.tgz",
+			"integrity": "sha512-GjbjXD4XXfN19D0LZNbmiCBUoDiRACsYHr0yaIbbn8aFsXjHZifcYqu/W5Er5X2X990WjHXFrxarn5chzItorQ==",
 			"cpu": [
 				"ppc64"
 			],
@@ -929,9 +1070,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-s390x-gnu": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-s390x-gnu/-/binding-linux-s390x-gnu-1.1.5.tgz",
-			"integrity": "sha512-gLYb4BIadlfTOYT5gO503n8zQjXflgzpD0FcyKh0Mzx3rqCZKnHoJWV9xe1KXUJ5lx2JfcSHr/mhzS0PC/McAA==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-s390x-gnu/-/binding-linux-s390x-gnu-1.2.4.tgz",
+			"integrity": "sha512-p5WR0NOwaRmJ/B1b6IjEFLLivwEsf3PrdBIhRbhTCQisbo2SvHHpG4ELB/+FgQNnB88LTOF86upmJmbvZdQ2lw==",
 			"cpu": [
 				"s390x"
 			],
@@ -949,9 +1090,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-x64-gnu": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-gnu/-/binding-linux-x64-gnu-1.1.5.tgz",
-			"integrity": "sha512-FjcpEKUyJygHgs1o50VYNvkt5+7Le/VEdYt0AkRpkL33MnyQfwr8l5mXwMmfmTbyMPr5vJLC+8/Gd9gXnwU1QQ==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-gnu/-/binding-linux-x64-gnu-1.2.4.tgz",
+			"integrity": "sha512-4/GyVjmhR+Tc6HLJvwc1sOhPqAZtySiSMesOZyX6JQ5XBxoTDEMKQzvo07NIK6nTon/SivlZqvhzvuVBNQhObQ==",
 			"cpu": [
 				"x64"
 			],
@@ -969,9 +1110,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-linux-x64-musl": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-musl/-/binding-linux-x64-musl-1.1.5.tgz",
-			"integrity": "sha512-Me+PfPI2TMeOQk0gYWfLQZtTktrmzbr8cDboqX83XKc7UrgAi55gF+2dUkWdxd19n55Essp2yeca+O9N5rBxHg==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-linux-x64-musl/-/binding-linux-x64-musl-1.2.4.tgz",
+			"integrity": "sha512-l9eeLsCNvPpmSXUej0etw/J1eqV0Jj1D5G/xG6YTijmE6dkv6E2QezgWbTfQk63v952DPqrjOCoiqxq7Bw0YUQ==",
 			"cpu": [
 				"x64"
 			],
@@ -989,9 +1130,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-openharmony-arm64": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-openharmony-arm64/-/binding-openharmony-arm64-1.1.5.tgz",
-			"integrity": "sha512-yc5WrLzXks6zCQfn9Oxr8pORKyl/pF+QjHmW/Qx3qu0oyrrNC+y2JLTU1E2rcWYAmzlnqngWXHQjy51VzW70Vw==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-openharmony-arm64/-/binding-openharmony-arm64-1.2.4.tgz",
+			"integrity": "sha512-e0F355MSTMm3+UOqtV3L24gFUp2N5m1f8L/7d56deik6va+AXdrt9F8LbzGpeWGWRbZEDq4m8NVnJDeBtf9DZg==",
 			"cpu": [
 				"arm64"
 			],
@@ -1005,63 +1146,10 @@
 				"node": "^20.19.0 || >=22.12.0"
 			}
 		},
-		"node_modules/@rolldown/binding-wasm32-wasi": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-wasm32-wasi/-/binding-wasm32-wasi-1.1.5.tgz",
-			"integrity": "sha512-VbQGPX2b4r48TAMIM2cjgluIM1HYutm4pcTEJsle7iEP7sB1dFqtPLBVbdLAZCxy1txCcPxf4QFf4v8uvltPqA==",
-			"cpu": [
-				"wasm32"
-			],
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"dependencies": {
-				"@emnapi/core": "1.11.1",
-				"@emnapi/runtime": "1.11.1",
-				"@napi-rs/wasm-runtime": "^1.1.6"
-			},
-			"engines": {
-				"node": "^20.19.0 || >=22.12.0"
-			}
-		},
-		"node_modules/@rolldown/binding-wasm32-wasi/node_modules/@emnapi/core": {
-			"version": "1.11.1",
-			"resolved": "https://registry.npmjs.org/@emnapi/core/-/core-1.11.1.tgz",
-			"integrity": "sha512-RSvbQmHzdKzNsLYa/wHrbc3KN4sYLKAdPZxqiM2HATqv/SBk2/ENSHpvXGaLOMcsAyz0poEGqkmmKYG3OWiJEQ==",
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"dependencies": {
-				"@emnapi/wasi-threads": "1.2.2",
-				"tslib": "^2.4.0"
-			}
-		},
-		"node_modules/@rolldown/binding-wasm32-wasi/node_modules/@emnapi/runtime": {
-			"version": "1.11.1",
-			"resolved": "https://registry.npmjs.org/@emnapi/runtime/-/runtime-1.11.1.tgz",
-			"integrity": "sha512-vgj7R3y3Wgx24IQaGPA/R6YFXLHVMOZ0uVEyIQPaWs+rd1AzfEMXlAC22FYwO1XkKR6NPsq7mUandH8oIRdZFw==",
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"dependencies": {
-				"tslib": "^2.4.0"
-			}
-		},
-		"node_modules/@rolldown/binding-wasm32-wasi/node_modules/@emnapi/wasi-threads": {
-			"version": "1.2.2",
-			"resolved": "https://registry.npmjs.org/@emnapi/wasi-threads/-/wasi-threads-1.2.2.tgz",
-			"integrity": "sha512-c95qOXkHdydNKhscBTebqEC1CVAZpyqOfVfBzQ1qgzyl3gfeldUjIggDbIZgDKsHLgnsM+igH7TJ/eAasaVuMA==",
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"dependencies": {
-				"tslib": "^2.4.0"
-			}
-		},
 		"node_modules/@rolldown/binding-win32-arm64-msvc": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-win32-arm64-msvc/-/binding-win32-arm64-msvc-1.1.5.tgz",
-			"integrity": "sha512-gHv82k63z4qpV5+Q1y/12KrK0ltWBukVDI8nZcbT7Tt/ZlOIVwppazneq0F93oDxTo3IgAMEDIoQh3E2n6mVsw==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-win32-arm64-msvc/-/binding-win32-arm64-msvc-1.2.4.tgz",
+			"integrity": "sha512-AWLi0uBRYh6QlE7OKhiz+phZC0qwtij2QZmhmOdsLdFn64m7oMpooE9ICE3lhm9xMb4SpDo2WbHcxX1iFLFtqw==",
 			"cpu": [
 				"arm64"
 			],
@@ -1076,9 +1164,9 @@
 			}
 		},
 		"node_modules/@rolldown/binding-win32-x64-msvc": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/@rolldown/binding-win32-x64-msvc/-/binding-win32-x64-msvc-1.1.5.tgz",
-			"integrity": "sha512-tTZuDBPw85tEN5PQi1pnEBzDy0Z49HtScLAbD5t6hyeU92A95pRWaSMw1GZZi/RwgSgUIl0xrSlXIT/9QzvYSA==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/@rolldown/binding-win32-x64-msvc/-/binding-win32-x64-msvc-1.2.4.tgz",
+			"integrity": "sha512-UwSDJOg3dqCAejWdxclJjCsh3Qq4vLYMDxmyHqo1btz3stK2VqgwNd3mm5tuIwzSlGIQ/1H9Hr+Zn09mrezNqQ==",
 			"cpu": [
 				"x64"
 			],
@@ -1317,17 +1405,6 @@
 				"@tensorflow/tfjs-core": "4.22.0"
 			}
 		},
-		"node_modules/@tybys/wasm-util": {
-			"version": "0.10.3",
-			"resolved": "https://registry.npmjs.org/@tybys/wasm-util/-/wasm-util-0.10.3.tgz",
-			"integrity": "sha512-F3fo1MYrRJYL3zER0OUOmkutjr1Vp23m7OsSgp7nq4SP6OqX6C/56XFIPAl5bt3zaBRjmW7SGz3u/6LwFpYcOg==",
-			"dev": true,
-			"license": "MIT",
-			"optional": true,
-			"dependencies": {
-				"tslib": "^2.4.0"
-			}
-		},
 		"node_modules/@types/estree": {
 			"version": "1.0.9",
 			"resolved": "https://registry.npmjs.org/@types/estree/-/estree-1.0.9.tgz",
@@ -1544,9 +1621,9 @@
 			"license": "MIT"
 		},
 		"node_modules/brace-expansion": {
-			"version": "2.1.3",
-			"resolved": "https://registry.npmjs.org/brace-expansion/-/brace-expansion-2.1.3.tgz",
-			"integrity": "sha512-DRdx5neNsG/QXbniLFWi2YmC/68oeOOmKz6zOjVk6ZS1ZLXgLIKqVEc6hWsmkjBbgii0SwaBTcJ5XKj5gzY/4A==",
+			"version": "2.1.4",
+			"resolved": "https://registry.npmjs.org/brace-expansion/-/brace-expansion-2.1.4.tgz",
+			"integrity": "sha512-hGfVzPxthbf3+2yjg/RBs60cB0FhqBS/zvdV/4wn4/BmN0bNMMHPc4V/BbFieqf1TKAGGAHnY4eSjajCl0f2Xg==",
 			"dev": true,
 			"license": "MIT",
 			"dependencies": {
@@ -3011,9 +3088,9 @@
 			"license": "MIT"
 		},
 		"node_modules/nanoid": {
-			"version": "3.3.16",
-			"resolved": "https://registry.npmjs.org/nanoid/-/nanoid-3.3.16.tgz",
-			"integrity": "sha512-bzlKTyNJ7+LdGIIwy8ijFpIqEQIvafahV7eYykJ8Cvh42EdJeODoJ6gUJXpQJvej1BddH8OqTXZNE/KfbWAu8Q==",
+			"version": "3.3.18",
+			"resolved": "https://registry.npmjs.org/nanoid/-/nanoid-3.3.18.tgz",
+			"integrity": "sha512-DTg4MJbGMWkfi6VZFdNt2/caMbQy4Ou+Op/hJQvGEWcnVfoA1QA+xzRKAzw9jD6+GVOOeYr/mIcuDSdug6F6+w==",
 			"dev": true,
 			"funding": [
 				{
@@ -3233,6 +3310,38 @@
 			"dev": true,
 			"license": "MIT"
 		},
+		"node_modules/pdf-parse": {
+			"version": "2.4.5",
+			"resolved": "https://registry.npmjs.org/pdf-parse/-/pdf-parse-2.4.5.tgz",
+			"integrity": "sha512-mHU89HGh7v+4u2ubfnevJ03lmPgQ5WU4CxAVmTSh/sxVTEDYd1er/dKS/A6vg77NX47KTEoihq8jZBLr8Cxuwg==",
+			"license": "Apache-2.0",
+			"dependencies": {
+				"@napi-rs/canvas": "0.1.80",
+				"pdfjs-dist": "5.4.296"
+			},
+			"bin": {
+				"pdf-parse": "bin/cli.mjs"
+			},
+			"engines": {
+				"node": ">=20.16.0 <21 || >=22.3.0"
+			},
+			"funding": {
+				"type": "github",
+				"url": "https://github.com/sponsors/mehmet-kozan"
+			}
+		},
+		"node_modules/pdfjs-dist": {
+			"version": "5.4.296",
+			"resolved": "https://registry.npmjs.org/pdfjs-dist/-/pdfjs-dist-5.4.296.tgz",
+			"integrity": "sha512-DlOzet0HO7OEnmUmB6wWGJrrdvbyJKftI1bhMitK7O2N8W2gc757yyYBbINy9IDafXAV9wmKr9t7xsTaNKRG5Q==",
+			"license": "Apache-2.0",
+			"engines": {
+				"node": ">=20.16.0 || >=22.3.0"
+			},
+			"optionalDependencies": {
+				"@napi-rs/canvas": "^0.1.80"
+			}
+		},
 		"node_modules/picocolors": {
 			"version": "1.1.1",
 			"resolved": "https://registry.npmjs.org/picocolors/-/picocolors-1.1.1.tgz",
@@ -3272,9 +3381,9 @@
 			"license": "MIT"
 		},
 		"node_modules/postcss": {
-			"version": "8.5.25",
-			"resolved": "https://registry.npmjs.org/postcss/-/postcss-8.5.25.tgz",
-			"integrity": "sha512-DTPx3RWSSnWyzLxQnlH0rJP+EW5ekl16ZU4/psbIhA0e53kJfdgaN5vKM+xP7yJtXVu+nfdVFmlgFDEKAe4Pyw==",
+			"version": "8.5.26",
+			"resolved": "https://registry.npmjs.org/postcss/-/postcss-8.5.26.tgz",
+			"integrity": "sha512-u82N74LFzG8ca+dD8puPnplTXoGH4fTPpVGuIbt36G3qvNlkvfD0lEAZSxaly3KX8TS/L1A1gsCEmvKmBcVbkQ==",
 			"dev": true,
 			"funding": [
 				{
@@ -3292,7 +3401,7 @@
 			],
 			"license": "MIT",
 			"dependencies": {
-				"nanoid": "^3.3.16",
+				"nanoid": "^3.3.17",
 				"picocolors": "^1.1.1",
 				"source-map-js": "^1.2.1"
 			},
@@ -3434,13 +3543,13 @@
 			"license": "BSD-3-Clause"
 		},
 		"node_modules/rolldown": {
-			"version": "1.1.5",
-			"resolved": "https://registry.npmjs.org/rolldown/-/rolldown-1.1.5.tgz",
-			"integrity": "sha512-t9z29cJjXf/vxQ8dyhCSpt6H6aSwHTk8cT5I3iy6SMXuFpk5mB6PL6XfC8PCwrPTx93udwKUm9HRteAlTGBLiA==",
+			"version": "1.2.4",
+			"resolved": "https://registry.npmjs.org/rolldown/-/rolldown-1.2.4.tgz",
+			"integrity": "sha512-rSr7irW0K7QRWzjdJXqZowkcRdDtjRduh43rBltnVKd0VFq839l1lJoDvGJb6gl7+4rTTCrPWu+YfujUL8Ug7w==",
 			"dev": true,
 			"license": "MIT",
 			"dependencies": {
-				"@oxc-project/types": "=0.139.0",
+				"@oxc-project/types": "=0.144.0",
 				"@rolldown/pluginutils": "^1.0.0"
 			},
 			"bin": {
@@ -3450,21 +3559,20 @@
 				"node": "^20.19.0 || >=22.12.0"
 			},
 			"optionalDependencies": {
-				"@rolldown/binding-android-arm64": "1.1.5",
-				"@rolldown/binding-darwin-arm64": "1.1.5",
-				"@rolldown/binding-darwin-x64": "1.1.5",
-				"@rolldown/binding-freebsd-x64": "1.1.5",
-				"@rolldown/binding-linux-arm-gnueabihf": "1.1.5",
-				"@rolldown/binding-linux-arm64-gnu": "1.1.5",
-				"@rolldown/binding-linux-arm64-musl": "1.1.5",
-				"@rolldown/binding-linux-ppc64-gnu": "1.1.5",
-				"@rolldown/binding-linux-s390x-gnu": "1.1.5",
-				"@rolldown/binding-linux-x64-gnu": "1.1.5",
-				"@rolldown/binding-linux-x64-musl": "1.1.5",
-				"@rolldown/binding-openharmony-arm64": "1.1.5",
-				"@rolldown/binding-wasm32-wasi": "1.1.5",
-				"@rolldown/binding-win32-arm64-msvc": "1.1.5",
-				"@rolldown/binding-win32-x64-msvc": "1.1.5"
+				"@rolldown/binding-android-arm64": "1.2.4",
+				"@rolldown/binding-darwin-arm64": "1.2.4",
+				"@rolldown/binding-darwin-x64": "1.2.4",
+				"@rolldown/binding-freebsd-x64": "1.2.4",
+				"@rolldown/binding-linux-arm-gnueabihf": "1.2.4",
+				"@rolldown/binding-linux-arm64-gnu": "1.2.4",
+				"@rolldown/binding-linux-arm64-musl": "1.2.4",
+				"@rolldown/binding-linux-ppc64-gnu": "1.2.4",
+				"@rolldown/binding-linux-s390x-gnu": "1.2.4",
+				"@rolldown/binding-linux-x64-gnu": "1.2.4",
+				"@rolldown/binding-linux-x64-musl": "1.2.4",
+				"@rolldown/binding-openharmony-arm64": "1.2.4",
+				"@rolldown/binding-win32-arm64-msvc": "1.2.4",
+				"@rolldown/binding-win32-x64-msvc": "1.2.4"
 			}
 		},
 		"node_modules/safe-buffer": {
@@ -4021,16 +4129,16 @@
 			}
 		},
 		"node_modules/vite": {
-			"version": "8.1.5",
-			"resolved": "https://registry.npmjs.org/vite/-/vite-8.1.5.tgz",
-			"integrity": "sha512-7ULLwsCdYx/nRyrpiEwvqb5TFHrMVZyBt+rg/OAXT7rgj/z+DtTDyKFeLAdDkubDVDKD8jOsndmy7m55XcfUsw==",
+			"version": "8.2.1",
+			"resolved": "https://registry.npmjs.org/vite/-/vite-8.2.1.tgz",
+			"integrity": "sha512-EU/eS7BH3XROHh2YnBefjM6DBKA6ZeMZEYQbj7NLWg5wHYlhB8B/Mayd5XsgWq+NFYccDOTemRpdETWR6Ka/lw==",
 			"dev": true,
 			"license": "MIT",
 			"dependencies": {
-				"lightningcss": "^1.32.0",
+				"lightningcss": "^1.33.0",
 				"picomatch": "^4.0.5",
-				"postcss": "^8.5.17",
-				"rolldown": "~1.1.5",
+				"postcss": "^8.5.25",
+				"rolldown": "~1.2.1",
 				"tinyglobby": "^0.2.17"
 			},
 			"bin": {
@@ -4047,7 +4155,7 @@
 			},
 			"peerDependencies": {
 				"@types/node": "^20.19.0 || >=22.12.0",
-				"@vitejs/devtools": "^0.3.0",
+				"@vitejs/devtools": "^0.4.0",
 				"esbuild": "^0.27.0 || ^0.28.0",
 				"jiti": ">=1.21.0",
 				"less": "^4.0.0",
@@ -4132,9 +4240,9 @@
 			"license": "MIT"
 		},
 		"node_modules/wasm-feature-detect": {
-			"version": "1.8.0",
-			"resolved": "https://registry.npmjs.org/wasm-feature-detect/-/wasm-feature-detect-1.8.0.tgz",
-			"integrity": "sha512-zksaLKM2fVlnB5jQQDqKXXwYHLQUVH9es+5TOOHwGOVJOCeRBCiPjwSg+3tN2AdTCzjgli4jijCH290kXb/zWQ==",
+			"version": "1.9.0",
+			"resolved": "https://registry.npmjs.org/wasm-feature-detect/-/wasm-feature-detect-1.9.0.tgz",
+			"integrity": "sha512-zonE+xlIIYtxPy++L24ow0hAD8CICb4+FgPyROd3buyXIqsJvUEDkBgfCCoXOd1Hu3DUr0GOfnPIdcGV+YpNaA==",
 			"license": "Apache-2.0"
 		},
 		"node_modules/webidl-conversions": {
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "1.5.0",
+	"version": "1.6.3",
 	"description": "AI Utility library",
 	"author": "Zak Timson",
 	"license": "MIT",
@@ -26,12 +26,13 @@
 	},
 	"dependencies": {
 		"@anthropic-ai/sdk": "^0.102.0",
-		"@tensorflow/tfjs": "^4.22.0",
 		"@huggingface/transformers": "^4.2.0",
+		"@tensorflow/tfjs": "^4.22.0",
 		"@ztimson/node-utils": "^1.0.7",
 		"@ztimson/utils": "^0.29.4",
 		"cheerio": "^1.2.0",
 		"openai": "^6.42.0",
+		"pdf-parse": "^2.4.5",
 		"tesseract.js": "^7.0.0"
 	},
 	"devDependencies": {
--- a/src/antrhopic.ts
+++ b/src/antrhopic.ts
@@ -24,6 +24,13 @@ export class Anthropic extends LLMProvider {
 		return client;
 	}

+	private toWireContent(content: any): any {
+		if(!Array.isArray(content)) return content;
+		return content.map(c => c.type === 'image'
+			? {type: 'image', source: {type: 'base64', media_type: c.mime, data: c.data}}
+			: {type: 'text', text: c.text});
+	}
+
 	/** Convert standard history -> Anthropic wire format */
 	private toWire(history: LLMMessage[]): any[] {
 		const wire: any[] = [];
@@ -34,7 +41,7 @@ export class Anthropic extends LLMProvider {
 					{role: 'user', content: [{type: 'tool_result', tool_use_id: h.id, is_error: !!h.error, content: h.error || h.content || ''}]}
 				);
 			} else {
-				wire.push({role: h.role, content: h.content});
+				wire.push({role: h.role, content: this.toWireContent(h.content)});
 			}
 		}
 		return wire;
--- a/src/llm.ts
+++ b/src/llm.ts
@@ -5,11 +5,16 @@ import {OpenAi} from './open-ai.ts';
 import {LLMProvider} from './provider.ts';
 import {AiTool, AiToolArg} from './tools.ts';
 import {fileURLToPath} from 'url';
-import {dirname, join} from 'path';
 import {spawn} from 'node:child_process';
 import {Memory, MemoryCache, MemoryManager, MemoryOptions, stripHeader} from './memory.ts';
+import {mkdtempSync} from 'node:fs';
+import fs from 'node:fs/promises';
+import {tmpdir} from 'node:os';
+import {dirname, join, basename, extname} from 'path';
+import { PDFParse } from 'pdf-parse';

 const MAX_AGENT_DEPTH = 5;
+const PDF_OCR_PAGE_THRESHOLD = 12; // above this many pages, OCR scanned pages instead of feeding images to the model

 export type AnthropicConfig = {proto: 'anthropic', token: string | string[]};
 export type OpenAiConfig = {proto: 'openai', host?: string, token: string | string[]};
@@ -27,11 +32,26 @@ export type Agent = {
 	agents?: string[] | null;
 }

+export type LLMFile = {
+	/** Path to file on disk */
+	path?: string;
+	/** File content: raw text, base64-encoded binary, or a Buffer */
+	content?: string | Buffer;
+	/** Original filename, used to infer type from extension */
+	name?: string;
+	/** Mime type override, inferred from extension if omitted */
+	mime?: string;
+	/** @internal set once extraction has run, skips re-processing next turn */
+	extracted?: boolean;
+};
+
 export type LLMMessage = {
 	/** Message originator */
 	role: 'assistant' | 'system' | 'user';
 	/** Message content */
 	content: string | any;
+	/** Files attached to request */
+	files?: LLMFile[];
 	/** Timestamp */
 	timestamp?: number;
 	/** Response duration in ms */
@@ -88,6 +108,8 @@ export type LLMRequest = {
 	mcp?: McpServer[];
 	/** Subagents exposed as delegatable/wrapped tools */
 	agents?: Agent[];
+	/** Attach files to request */
+	files?: LLMFile[];
 	/** @internal recursion guard for nested agent delegation */
 	_agentDepth?: number;
 }
@@ -111,6 +133,11 @@ export type Skill = {
 }

 class LLM {
+	private static AUDIO_EXT = ['wav','mp3','m4a','flac','ogg','aac','wma'];
+	private static IMAGE_EXT = ['png','jpg','jpeg','bmp','gif','tiff','webp'];
+	private static TEXT_EXT = ['txt','md','csv','json','xml','html','js','ts','py','yaml','yml','log'];
+	private static PDF_EXT = ['pdf'];
+
 	private memoryManager!: MemoryManager;

 	defaultModel!: string;
@@ -126,6 +153,118 @@ class LLM {
 		this.memoryManager = new MemoryManager(this);
 	}

+	private async loadBuffer(file: LLMFile, asText: boolean): Promise<Buffer> {
+		if(file.path) return fs.readFile(file.path);
+		if(Buffer.isBuffer(file.content)) return file.content;
+		if(typeof file.content === 'string') return Buffer.from(file.content, asText ? 'utf-8' : 'base64');
+		throw new Error('No path or content provided');
+	}
+
+	private async writeTemp(name: string, buffer: Buffer): Promise<string> {
+		const path = join(mkdtempSync(join(tmpdir(), 'ai-file-')), name);
+		await fs.writeFile(path, buffer);
+		return path;
+	}
+
+	/**
+	 * Extract text from a PDF. Pages with no text layer (scanned/image-only) are handled as either:
+	 * - Rendered to images and returned alongside the text so the (vision-capable) model can read them directly
+	 * - OCR'd via Tesseract when the doc is too large to reasonably pass as images
+	 */
+	private async resolvePdf(buffer: Buffer): Promise<{text: string, images: {mime: string, data: string}[]}> {
+		const parser = new PDFParse({data: buffer});
+		try {
+			const {text, pages} = await parser.getText();
+			const scanned = (pages || []).filter(p => !p.text?.trim());
+			if(!scanned.length) return {text: text.trim() || '[Empty PDF]', images: []};
+			const total = pages.length;
+			const pageNums = scanned.map(p => p.num);
+			const {pages: shots} = await parser.getScreenshot({partial: pageNums});
+			if(total <= PDF_OCR_PAGE_THRESHOLD) {
+				return {
+					text: text.trim(),
+					images: shots.map(s => ({mime: 'image/png', data: Buffer.from(s.data).toString('base64')}))
+				};
+			}
+			const ocrText = await Promise.all(shots.map(async (s, i) => {
+				const path = await this.writeTemp(`page-${pageNums[i]}.png`, Buffer.from(s.data));
+				try {
+					return await this.ai.vision.ocr(path) || '';
+				} finally {
+					fs.rm(dirname(path), {recursive: true, force: true}).catch(() => {});
+				}
+			}));
+			return {text: [text.trim(), ...ocrText].filter(Boolean).join('\n\n'), images: []};
+		} finally {
+			await parser.destroy();
+		}
+	}
+
+	private async resolveFile(file: LLMFile): Promise<{text?: string, images?: {mime: string, data: string}[]}> {
+		const name = file.name || (file.path ? basename(file.path) : 'file');
+
+		// Already resolved on a previous turn, reuse cached text
+		if(file.extracted) return {text: `<file name="${name}">\n${file.content}\n</file>`};
+
+		const ext = extname(name).slice(1).toLowerCase();
+		const mime = file.mime || '';
+		const isAudio = mime.startsWith('audio/') || LLM.AUDIO_EXT.includes(ext);
+		const isImage = mime.startsWith('image/') || LLM.IMAGE_EXT.includes(ext);
+		const isPdf = mime === 'application/pdf' || LLM.PDF_EXT.includes(ext);
+		const isText = mime.startsWith('text/') || LLM.TEXT_EXT.includes(ext);
+
+		let tmpDir: string | null = null;
+		try {
+			if(isImage) {
+				const data = (await this.loadBuffer(file, false)).toString('base64');
+				return {images: [{mime: mime || `image/${ext === 'jpg' ? 'jpeg' : ext}`, data}]};
+			}
+
+			if(isPdf) {
+				const {text, images} = await this.resolvePdf(await this.loadBuffer(file, false));
+				// Only cache/skip re-processing when we didn't need to hand off images (OCR'd or fully text-based)
+				if(!images.length) {
+					file.content = text;
+					file.extracted = true;
+					delete file.path;
+				}
+				return {text: `<file name="${name}">\n${text || '[Scanned PDF - see attached page images]'}\n</file>`, images};
+			}
+
+			let text: string;
+			if(isAudio) {
+				let path = file.path;
+				if(!path) {
+					const buffer = await this.loadBuffer(file, false);
+					path = await this.writeTemp(name, buffer);
+					tmpDir = dirname(path);
+				}
+				text = await this.ai.audio.asr(path) || '';
+			} else if(isText) {
+				text = (await this.loadBuffer(file, true)).toString('utf-8');
+			} else {
+				text = typeof file.content === 'string' ? file.content : `[Binary file, unable to extract: ${name}]`;
+			}
+			file.content = text;
+			file.extracted = true;
+			delete file.path;
+
+			return {text: `<file name="${name}">\n${text}\n</file>`};
+		} catch(err: any) {
+			return {text: `<file name="${name}">Failed to process: ${err.message}</file>`};
+		} finally {
+			if(tmpDir) fs.rm(tmpDir, {recursive: true, force: true}).catch(() => {});
+		}
+	}
+
+	private async resolveFiles(files: LLMFile[]): Promise<{text: string, images: {mime: string, data: string}[]}> {
+		const resolved = await Promise.all(files.map(f => this.resolveFile(f)));
+		return {
+			text: resolved.filter(r => r.text).map(r => r.text).join('\n\n'),
+			images: resolved.flatMap(r => r.images || [])
+		};
+	}
+
 	private setupAgent(agents: Agent[] = [], allAgents: Agent[], history: LLMMessage[], aborts: (() => void)[], depth = 0, delegateState: {resp: string | null}): AiTool[] {
 		return agents.map(a => {
 			const toolName = `${a.delegate ? '' : 'sub'}agent_${snakeCase(a.name)}`;
@@ -272,7 +411,8 @@ ${a.system}`,
 			let tools: AiTool[] = options.tools || this.ai.options.llm?.tools || [];
 			const prompts: string[] = [];
 			let history = options.history || [];
-			if(message) history.push({role: 'user', content: message, timestamp: Date.now()});
+			const files = options.files || [];
+			if(message || files.length) history.push({role: 'user', content: message || '', timestamp: Date.now()});

 			// MCP
 			const mcp = options.mcp || this.ai.options?.llm?.mcp;
@@ -343,6 +483,20 @@ Linked: ${makeUnique([...r.links, ...r.backlinks]).join(', ')}

 			if(aborted) throw Object.assign(new Error('Aborted'), {name: 'AbortError'});

+			const lastMsg = history[history.length - 1];
+			if(files.length && lastMsg?.role === 'user') lastMsg.files = files;
+			const restores: {msg: LLMMessage, content: any}[] = [];
+			for(const msg of history) {
+				if(msg.role !== 'user' || !msg.files?.length) continue;
+				const {text, images} = await this.resolveFiles(msg.files);
+				if(!text && !images.length) continue;
+				restores.push({msg, content: msg.content});
+				const merged = text ? [msg.content, text].filter(Boolean).join('\n\n') : msg.content;
+				msg.content = images.length
+					? [...images.map(i => ({type: 'image', mime: i.mime, data: i.data})), {type: 'text', text: merged}]
+					: merged;
+			}
+
 			const toolTimings = new Map<string, {duration: number, tps: number}>();
 			tools = this.wrapToolTiming(tools, toolTimings);

@@ -352,6 +506,9 @@ Linked: ${makeUnique([...r.links, ...r.backlinks]).join(', ')}
 			request = this.models[m].ask('', {...options, tools, system: prompts.filter(Boolean).join('\n\n')});
 			let resp = await request;

+			// Strip the file injection shim
+			restores.forEach(({msg, content}) => msg.content = content);
+
 			// Capture meta (duration / tps)
 			for(const h of history) {
 				if(h.role === 'tool' && toolTimings.has(h.id)) Object.assign(h, toolTimings.get(h.id));
--- a/src/memory.ts
+++ b/src/memory.ts
@@ -2,9 +2,10 @@ import {MemoryNode, rebuildGraph} from './helpers.ts';
 import {LLMRequest, LLMMessage} from './llm.ts';
 import {AiTool} from './tools.ts';
 import {KDPoint, KDTree} from './kd-tree.ts';
+import {escapeRegex} from '@ztimson/utils';

-const FACTS_HEADING = '## Facts';
-
+const MERGE_THRESHOLD = 0.88;
+const PENDING_HEADING = '## Pending';
 const GENERIC_TEMPLATE = `# {{Title}}

 ## Summary
@@ -32,6 +33,11 @@ type FactBucket = {
 	facts: string[];
 }

+type FactAgentResult = {
+	buckets: FactBucket[];
+	journal: string;
+}
+
 function dedupeFacts(facts: string[]): string[] {
 	const seen = new Map<string, string>();
 	for (const f of facts) {
@@ -251,14 +257,13 @@ ${m.content}
 		return new MemoryAccessor(memories);
 	}

-	private appendFacts(node: Memory, facts: string[]): void {
+	private stage(node: Memory, block: string): void {
 		this.ensureDoc(node);
 		const body = stripHeader(node.content);
-		const bullets = facts.map(f => `- ${f}`).join('\n');
-		const idx = body.indexOf(FACTS_HEADING);
+		const idx = body.indexOf(PENDING_HEADING);
 		const newBody = idx === -1
-			? `${body.trimEnd()}\n\n${FACTS_HEADING}\n${bullets}\n`
-			: `${body.slice(0, idx + FACTS_HEADING.length)}\n${bullets}${body.slice(idx + FACTS_HEADING.length)}`;
+			? `${body.trimEnd()}\n\n${PENDING_HEADING}\n${block}\n`
+			: `${body.slice(0, idx + PENDING_HEADING.length)}\n${block}${body.slice(idx + PENDING_HEADING.length)}`;
 		node.content = this.touchHeader(node, newBody);
 	}

@@ -268,16 +273,30 @@ ${m.content}
 		node.content = this.touchHeader(node, `# ${title}\n`);
 	}

-	private async factAgent(conversation: string, store: MemoryAccessor, options: LLMRequest, weekKey: string): Promise<FactBucket[]> {
+	private sanitizeDescription(text: string): string {
+		return (text ?? '').replace(/\s+/g, ' ').trim().slice(0, 240);
+	}
+
+	private relink(memories: Memory[], from: string, to: string): void {
+		const pattern = new RegExp(`\\[\\[${escapeRegex(from)}\\]\\]`, 'g');
+		for (const m of memories) if (pattern.test(m.content)) m.content = m.content.replace(pattern, `[[${to}]]`);
+	}
+
+	private async factAgent(conversation: string, store: MemoryAccessor, options: LLMRequest): Promise<FactAgentResult> {
 		const ghosts = store.ghosts();

 		const response = await this.llm.ask(conversation, {
 			model: options.model,
 			temperature: 0.2,
 			system: `You are a fact extractor to build obsidian knowledge vaults.
-Analyze this conversation and extract facts worth remembering long-term.
+Analyze this conversation and produce two things: a journal recap and any long-term facts worth filing.

-Rules:
+Journal recap rules:
+- Write a short day-to-day recap of this conversation: what was discussed/worked on, decisions still in flux, the user's state/mood if mentioned, general context
+- Write something whenever there was any meaningful activity, even if nothing below qualifies as a hard fact
+- Leave it empty only for trivial/empty exchanges (greetings, pleasantries, nothing happened)
+
+Fact bucket rules:
 - Always extract facts that the user explicitly told you to remember
 - ONLY extract current facts the USER explicitly stated about themselves, their work, projects or decisions that were MADE during this conversation
 - DO NOT extract greetings, pleasantries, or generic exchanges
@@ -286,19 +305,20 @@ Rules:
 - If nothing worth remembering was said, return an empty buckets array

 When extracting facts, you MUST also decide the exact destination path:
- Reuse node names (including ghost) as much as possible IF the facts belongs there
+- Reuse node names (including ghost) as much as possible IF the fact belongs there
 - All information primarily about the user should go under "People/User"
- When required, create a new path following collection/subject format (e.g., People/Sarah, Projects/Oxide) — you are not limited to any fixed list of collections, use whatever fits
- For journal entries, use "Journal"
+- NEVER use a generic label like "Technical Issue", "Technical Decision", "Technical Requirement", "Bug", "Feature Request" as a subject — these fragment the vault. Every fact belongs to a concrete subject: a person, project, or system.
+- A bug report, its investigation, and its eventual fix are the SAME topic — reuse the one node for the subject, don't fork a new one per event
+- When required, create a new path following collection/subject format (e.g., People/Sarah, Projects/Oxide), named after the subject, not the event — you are not limited to any fixed list of collections

 Available nodes:
- Journal
-${this.listNodes(store.list).filter(n => !n.name.includes('Journal')).map(n => `- ${n.name}: ${n.description}`).join('\n') || 'None yet.'}
+${this.listNodes(store.list).map(n => `- ${n.name}: ${n.description}`).join('\n') || 'None yet.'}
 ${ghosts.length ? `${ghosts.map(g => `- ${g}: (Ghost)`).join('\n')}` : ''}`,
 			schema: {
+				journal: {type: 'string', description: 'Short day-to-day recap of this conversation, see rules above. Empty string if nothing happened.', required: false},
 				buckets: {type: 'array', description: 'Groups of facts to remember, each assigned to a different node. Return an empty array if there is nothing worth storing in an obsidian vault', items: {
 						type: 'object', items: {
-							subject: {type: 'string', description: 'Exact existing node name OR new path (e.g. "People/Sarah", "Projects/Oxide"), or "Journal"', required: true},
+							subject: {type: 'string', description: 'Exact existing node name OR new path (e.g. "People/Sarah", "Projects/Oxide")', required: true},
 							facts: {
 								type: 'array',
 								description: 'Facts to store at this destination',
@@ -311,15 +331,17 @@ ${ghosts.length ? `${ghosts.map(g => `- ${g}: (Ghost)`).join('\n')}` : ''}`,
 		});

 		const buckets = new Map<string, string[]>();
-		for(const bucket of response.buckets ?? []) {
-			const subject = bucket.subject.trim().toLowerCase() === 'journal'
-				? `Journal/${weekKey}` : bucket.subject.trim();
+		for (const bucket of response.buckets ?? []) {
+			const subject = bucket.subject.trim();
 			const facts = buckets.get(subject) ?? [];
 			facts.push(...dedupeFacts(bucket.facts));
 			buckets.set(subject, facts);
 		}

-		return buckets.entries().toArray().map(([subject, facts]) => ({subject, facts}));
+		return {
+			buckets: buckets.entries().toArray().map(([subject, facts]) => ({subject, facts})),
+			journal: (response.journal ?? '').trim(),
+		};
 	}

 	private getWeekMonday(date: Date = new Date()): string {
@@ -334,6 +356,34 @@ ${ghosts.length ? `${ghosts.map(g => `- ${g}: (Ghost)`).join('\n')}` : ''}`,
 		return memories.map(m => ({name: m.name, description: m.description}));
 	}

+	/** Find the nearest node above the similarity threshold and fold the smaller/less-connected one into
+	 * the other. Journals are exempt — they're partitioned by date, not topic, and merging across weeks
+	 * would wreck the timeline. Returns 'merged' if `node` absorbed another (caller should re-run the doc
+	 * agent), 'absorbed' if `node` itself got folded away (caller should stop touching it), or null. */
+	private async checkMerge(node: Memory, memories: Memory[] | MemoryCache, options: LLMRequest, threshold = MERGE_THRESHOLD): Promise<'merged' | 'absorbed' | null> {
+		if (!node.embedding?.length || node.name.startsWith('Journal/')) return null;
+		const store = this.access(memories);
+
+		let closest: Memory | null = null, closestDist = Infinity;
+		for (const other of store.list) {
+			if (other.name === node.name || other.name.startsWith('Journal/') || !other.embedding?.length) continue;
+			const d = cosineDistance(node.embedding, other.embedding);
+			if (d < closestDist) { closestDist = d; closest = other; }
+		}
+		if (!closest || closestDist > threshold) return null;
+
+		const weight = (m: Memory) => m.content.length + (m.links.length + m.backlinks.length) * 200;
+		const [primary, secondary] = weight(node) >= weight(closest) ? [node, closest] : [closest, node];
+
+		this.stage(primary, `### Merged from [[${secondary.name}]]\n\n${stripHeader(secondary.content)}`);
+		this.relink(store.list, secondary.name, primary.name);
+		store.forget(secondary.name);
+
+		if (primary === node) return 'merged';
+		this.reconcile(primary, memories, options).catch(() => {});
+		return 'absorbed';
+	}
+
 	private reconcile(node: Memory, memories: Memory[] | MemoryCache, options: LLMRequest): Promise<void> {
 		const key = node.name;
 		const existing = this.queues.get(key);
@@ -350,6 +400,9 @@ ${ghosts.length ? `${ghosts.map(g => `- ${g}: (Ghost)`).join('\n')}` : ''}`,
 			do {
 				entry.dirty = false;
 				await this.docAgent(node, store.list, options, entry);
+				const merge = await this.checkMerge(node, memories, options);
+				if (merge === 'absorbed') break;
+				if (merge === 'merged') entry.dirty = true;
 			} while (entry.dirty);
 		})().finally(() => {
 			this.queues.delete(key);
@@ -360,6 +413,7 @@ ${ghosts.length ? `${ghosts.map(g => `- ${g}: (Ghost)`).join('\n')}` : ''}`,

 	private async docAgent(node: Memory, memories: Memory[], options: LLMRequest, entry: {request: {abort?: () => void} | null}): Promise<void> {
 		const currentBody = stripHeader(node.content);
+		const isJournal = node.name.startsWith('Journal/');
 		let update;
 		try {
 			for (let i = 0; i < 2 && !update?.content; i++) {
@@ -367,24 +421,27 @@ ${ghosts.length ? `${ghosts.map(g => `- ${g}: (Ghost)`).join('\n')}` : ''}`,
 					model: options.model,
 					temperature: 0.3,
 					schema: {
-						description: {type: 'string', description: 'One-line description of what this document covers, no formatting or emojis', required: true},
+						description: {type: 'string', description: 'One factual sentence describing what this document contains. Never a thought process, reasoning, or meta-commentary about the task.', required: true},
 						content: {type: 'string', description: 'Rewritten document body in markdown, without the frontmatter block', required: true},
 					},
 					system: `You are a knowledge base editor maintaining one document in an Obsidian-style vault.

-If the document has a "${FACTS_HEADING}" section, integrate every bullet under it into the appropriate part of the document, then remove the "${FACTS_HEADING}" section entirely. If there is no such section, just tidy the document per the rules below.
+If the document has a "${PENDING_HEADING}" section, it holds new material to integrate — bullet facts, a journal entry, or an entire document merged in from elsewhere (marked "### Merged from [[X]]" or "### {date}"). Fold all of it into the appropriate part of the document, resolving overlap and duplication, then remove the "${PENDING_HEADING}" section entirely. If there is no such section, just tidy the document per the rules below.

-Structure: follow this generic shape loosely, adapting section names/order to what the content actually needs (e.g. journal-style docs may want a timeline instead of "Details"):
+Structure: follow this generic shape loosely, adapting section names/order to what the content actually needs:
 \`\`\`markdown
 ${GENERIC_TEMPLATE}
 \`\`\`

+${isJournal
+						? `This document is a journal — a chronological day-to-day log. Keep entries ordered as a timeline under "### {date}" headings. NEVER delete or "resolve away" older entries in favor of newer ones — history stays, only clean up grammar/clarity within an entry.`
+						: `Resolve contradictions: newer facts always win — delete the outdated statement entirely, never keep both.`}
+
 Formatting rules:
 - Use Obsidian-style markdown: # headings, **bold** for emphasis, bullet & numbered lists for grouped 1D data, tables for 2D data
 - Link related concepts with [[WikiLink]] notation using full paths like [[People/Sarah]] or [[Projects/Website]]
 - Create links for specific entities (person, place, project, program) and abstract concepts, but skip generics (car, red, dog)
 - Keep the document concise, factual, and human-readable
- Resolve contradictions: newer facts always win — delete the outdated statement entirely, never keep both
 - Do not add frontmatter blocks, filler, preamble, or AI commentary

 Other nodes in the vault (link to these instead of duplicating their content):
@@ -406,7 +463,7 @@ ${currentBody}
 		}

 		if (!update?.content) return;
-		node.description = node.name !== 'People/User' ? update.description : 'All information about the current user';
+		node.description = node.name !== 'People/User' ? this.sanitizeDescription(update.description) : 'All information about the current user';
 		node.content = this.touchHeader(node, update.content);
 		const [e] = await this.llm.embedding(node.content);
 		if (e) node.embedding = e.embedding;
@@ -419,7 +476,11 @@ ${currentBody}
 		for (const line of match[1].split('\n')) {
 			const i = line.indexOf(':');
 			if (i === -1) continue;
-			fm.set(line.slice(0, i).trim(), line.slice(i + 1).trim());
+			const key = line.slice(0, i).trim();
+			const raw = line.slice(i + 1).trim();
+			let value = raw;
+			try { value = JSON.parse(raw); } catch { /* legacy unquoted value, keep raw */ }
+			fm.set(key, value);
 		}
 		return {fm, body: match[2]};
 	}
@@ -433,7 +494,7 @@ ${currentBody}
 	}

 	private writeFrontmatter(fm: Map<string, string>, body: string): string {
-		const lines = [...fm.entries()].map(([k, v]) => `${k}: ${v}`);
+		const lines = [...fm.entries()].map(([k, v]) => `${k}: ${JSON.stringify(String(v).replace(/\s+/g, ' ').trim())}`);
 		return `---\n${lines.join('\n')}\n---\n\n${body.trimStart()}`;
 	}

@@ -498,20 +559,34 @@ ${currentBody}
 		history.push(pending);

 		const store = this.access(memories);
-		const buckets = await this.factAgent(conversation, store, options, this.getWeekMonday());
+		const {buckets, journal} = await this.factAgent(conversation, store, options);
 		const touched: Memory[] = [];

+		if (journal) {
+			const journalName = `Journal/${this.getWeekMonday()}`;
+			let jnode = store.find(journalName);
+			if (!jnode) {
+				jnode = {name: journalName, description: '', content: '', embedding: [], links: [], backlinks: []};
+				store.list.push(jnode);
+			}
+			this.stage(jnode, `### ${new Date().toISOString().slice(0, 10)}\n${journal}`);
+			touched.push(jnode);
+		}
+
 		for (const {subject, facts} of buckets) {
 			let node = store.find(subject);
 			if (!node) {
 				node = {name: subject, description: '', content: '', embedding: [], links: [], backlinks: []};
 				store.list.push(node);
 			}
-			this.appendFacts(node, facts);
+			this.stage(node, facts.map(f => `- ${f}`).join('\n'));
+			touched.push(node);
+		}
+
+		for (const node of touched) {
 			const [e] = await this.llm.embedding(node.content);
 			if (e) node.embedding = e.embedding;
 			this.touch(node.name);
-			touched.push(node);
 		}

 		if (touched.length) {
@@ -526,9 +601,9 @@ ${currentBody}
 		return touched;
 	}

-	async reconcileVault(memories: Memory[] | MemoryCache, options: LLMRequest, scope: 'touched' | 'all' = 'touched'): Promise<void> {
+	async reconcileAll(memories: Memory[] | MemoryCache, options: LLMRequest, scope: 'touched' | 'all' = 'touched'): Promise<void> {
 		const store = this.access(memories);
-		const targets = scope === 'all' ? store.list : store.list.filter(m => m.content.includes(FACTS_HEADING));
+		const targets = scope === 'all' ? store.list : store.list.filter(m => m.content.includes(PENDING_HEADING));
 		await Promise.all(targets.map(node => this.reconcile(node, memories, options)));
 		store.commit();
 	}
--- a/src/open-ai.ts
+++ b/src/open-ai.ts
@@ -25,6 +25,13 @@ export class OpenAi extends LLMProvider {
 		return client;
 	}

+	private toWireContent(content: any): any {
+		if(!Array.isArray(content)) return content;
+		return content.map(c => c.type === 'image'
+			? {type: 'image_url', image_url: {url: `data:${c.mime};base64,${c.data}`}}
+			: {type: 'text', text: c.text});
+	}
+
 	/** Convert standard history -> OpenAI wire format */
 	private toWire(history: LLMMessage[], system?: string): any[] {
 		const wire: any[] = [];
@@ -41,7 +48,7 @@ export class OpenAi extends LLMProvider {
 					content: h.error || h.content || '',
 				});
 			} else {
-				wire.push({role: h.role, content: h.content});
+				wire.push({role: h.role, content: this.toWireContent(h.content)});
 			}
 		}
 		return wire;
Author	SHA1	Message	Date
ztimson	08a351e028	Better memory management All checks were successful Publish Library / Build NPM Project (push) Successful in 59s Details Publish Library / Tag Version (push) Successful in 11s Details	2026-08-24 14:42:10 -04:00
ztimson	85c01d3ef1	Added official file support All checks were successful Publish Library / Build NPM Project (push) Successful in 30s Details Publish Library / Tag Version (push) Successful in 10s Details	2026-08-17 15:50:48 -04:00
ztimson	5826573d5c	Added official file support All checks were successful Publish Library / Build NPM Project (push) Successful in 50s Details Publish Library / Tag Version (push) Successful in 13s Details	2026-08-17 15:16:32 -04:00
ztimson	797a40a566	Added official file support All checks were successful Publish Library / Build NPM Project (push) Successful in 58s Details Publish Library / Tag Version (push) Successful in 13s Details	2026-08-16 15:40:50 -04:00