red-candle 1.6.1 → 1.7.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 00c0870d599db76ba556ab880f69219bc32d72698c3570cd308df3923b1332f9
-  data.tar.gz: bffbd02fa1fe11813a304ed16771e6a72ee01f85ccb445eb67c5da6bf696c670
+  metadata.gz: 3d1b83311f0ad99adaffb886efda12291e9499ba112742e471e7b1eace390cee
+  data.tar.gz: 6156968c937204767fda21adfc6069e55a9571018f064ac56821d61472cc9cc1
 SHA512:
-  metadata.gz: c9ab39f01e4777c3d9906592cf62b8b1d2d03fd04be4ba82d8ddee837678ac477be88b1a0a48fa898e443f7469681e26f25297d4d310c39fb5222137d3b17d87
-  data.tar.gz: 22fe87f0fb64754a0a19bcc632dff72b0cc43838d73797aa991c961b65c9ecb7d9bd7aacbde51baf4063011081f4111a64f53726b51c87aca3e48214304e0119
+  metadata.gz: 57abc7d285ebb3c67c438d861563247ec1130b03bee5fab418670bd926de58ad2bc06b2328732c362f46a4f8a8ccb7d85d7bd0e4b7a384838e290dadb7ed2be4
+  data.tar.gz: 2f10da2df023629d59b2c2454efa886a87e0738d99801e6bb25b951431856b083780167190a6ba735c3c51cb8d170ed16a9c163f19aaf44fc8b95a72f8ea4b5e

data/Cargo.lock

@@ -46,6 +46,24 @@ dependencies = [
  "memchr",
 ]
 
+[[package]]
+name = "aligned"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ee4508988c62edf04abd8d92897fca0c2995d907ce1dfeaf369dac3716a40685"
+dependencies = [
+ "as-slice",
+]
+
+[[package]]
+name = "aligned-vec"
+version = "0.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc890384c8602f339876ded803c97ad529f3842aba97f6392b3dba0dd171769b"
+dependencies = [
+ "equator",
+]
+
 [[package]]
 name = "allocator-api2"
 version = "0.2.21"
@@ -67,6 +85,38 @@ version = "1.0.99"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b0674a1ddeecb70197781e945de4b3b8ffb61fa939a5597bcf48503737663100"
 
+[[package]]
+name = "arbitrary"
+version = "1.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c3d036a3c4ab069c7b410a2ce876bd74808d2d0888a82667669f8e783a898bf1"
+
+[[package]]
+name = "arg_enum_proc_macro"
+version = "0.3.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0ae92a5119aa49cdbcf6b9f893fe4e1d98b04ccbf82ee0584ad948a44a734dea"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "arrayvec"
+version = "0.7.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7c02d123df017efcdfbd739ef81735b36c5ba83ec3c59c80a9d7ecc718f92e50"
+
+[[package]]
+name = "as-slice"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "516b6b4f0e40d50dcda9365d53964ec74560ad4284da2e7fc97122cd83174516"
+dependencies = [
+ "stable_deref_trait",
+]
+
 [[package]]
 name = "atomic-waker"
 version = "1.1.2"
@@ -79,6 +129,49 @@ version = "1.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c08606f8c3cbf4ce6ec8e28fb0014a2c086708fe954eaa885384a6165172e7e8"
 
+[[package]]
+name = "av-scenechange"
+version = "0.14.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0f321d77c20e19b92c39e7471cf986812cbb46659d2af674adc4331ef3f18394"
+dependencies = [
+ "aligned",
+ "anyhow",
+ "arg_enum_proc_macro",
+ "arrayvec",
+ "log",
+ "num-rational",
+ "num-traits",
+ "pastey",
+ "rayon",
+ "thiserror 2.0.16",
+ "v_frame",
+ "y4m",
+]
+
+[[package]]
+name = "av1-grain"
+version = "0.2.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8cfddb07216410377231960af4fcab838eaa12e013417781b78bd95ee22077f8"
+dependencies = [
+ "anyhow",
+ "arrayvec",
+ "log",
+ "nom 8.0.0",
+ "num-rational",
+ "v_frame",
+]
+
+[[package]]
+name = "avif-serialize"
+version = "0.8.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "375082f007bd67184fb9c0374614b29f9aaa604ec301635f72338bb65386a53d"
+dependencies = [
+ "arrayvec",
+]
+
 [[package]]
 name = "aws-lc-rs"
 version = "1.14.0"
@@ -221,6 +314,12 @@ version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5e764a1d40d510daf35e07be9eb06e75770908c27d411ee6c92109c9840eaaf7"
 
+[[package]]
+name = "bit_field"
+version = "0.10.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e4b40c7323adcfc0a41c4b88143ed58346ff65a288fc144329c5c45e05d70c6"
+
 [[package]]
 name = "bitflags"
 version = "1.3.2"
@@ -233,6 +332,15 @@ version = "2.9.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2261d10cca569e4643e526d8dc2e62e433cc8aba21ab764233731f8d369bf394"
 
+[[package]]
+name = "bitstream-io"
+version = "4.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "60d4bd9d1db2c6bdf285e223a7fa369d5ce98ec767dec949c6ca62863ce61757"
+dependencies = [
+ "core2",
+]
+
 [[package]]
 name = "block"
 version = "0.1.6"
@@ -257,6 +365,12 @@ dependencies = [
  "objc2",
 ]
 
+[[package]]
+name = "built"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f4ad8f11f288f48ca24471bbd51ac257aaeaaa07adae295591266b792902ae64"
+
 [[package]]
 name = "bumpalo"
 version = "3.19.0"
@@ -289,6 +403,12 @@ version = "1.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
 
+[[package]]
+name = "byteorder-lite"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f1fe948ff07f4bd06c30984e69f5b4899c516a3ef74f34df92a2df2ab535495"
+
 [[package]]
 name = "bytes"
 version = "1.11.1"
@@ -304,6 +424,7 @@ dependencies = [
  "candle-transformers",
  "half",
  "hf-hub",
+ "image",
  "magnus",
  "outlines-core",
  "rand 0.8.5",
@@ -445,7 +566,7 @@ version = "0.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6fac387a98bb7c37292057cffc56d62ecb629900026402633ae9160df93a8766"
 dependencies = [
- "nom",
+ "nom 7.1.3",
 ]
 
 [[package]]
@@ -493,6 +614,12 @@ dependencies = [
  "cc",
 ]
 
+[[package]]
+name = "color_quant"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3d7b894f5411737b7867f4827955924d7c254fc9f4d91a6aad6b097804b1018b"
+
 [[package]]
 name = "compact_str"
 version = "0.9.0"
@@ -548,6 +675,15 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "core2"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b49ba7ef1ad6107f8824dbe97de947cbaac53c44e7f9756a1fba0d37c1eec505"
+dependencies = [
+ "memchr",
+]
+
 [[package]]
 name = "cpufeatures"
 version = "0.2.17"
@@ -812,6 +948,26 @@ dependencies = [
  "syn",
 ]
 
+[[package]]
+name = "equator"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4711b213838dfee0117e3be6ac926007d7f433d7bbe33595975d4190cb07e6fc"
+dependencies = [
+ "equator-macro",
+]
+
+[[package]]
+name = "equator-macro"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "44f23cf4b44bfce11a86ace86f8a73ffdec849c9fd00a386a53d278bd9e81fb3"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "equivalent"
 version = "1.0.2"
@@ -837,6 +993,21 @@ dependencies = [
  "cc",
 ]
 
+[[package]]
+name = "exr"
+version = "1.74.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4300e043a56aa2cb633c01af81ca8f699a321879a7854d3896a0ba89056363be"
+dependencies = [
+ "bit_field",
+ "half",
+ "lebe",
+ "miniz_oxide",
+ "rayon-core",
+ "smallvec",
+ "zune-inflate",
+]
+
 [[package]]
 name = "fancy-regex"
 version = "0.14.0"
@@ -865,6 +1036,35 @@ version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
 
+[[package]]
+name = "fax"
+version = "0.2.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f05de7d48f37cd6730705cbca900770cab77a89f413d23e100ad7fad7795a0ab"
+dependencies = [
+ "fax_derive",
+]
+
+[[package]]
+name = "fax_derive"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a0aca10fb742cb43f9e7bb8467c91aa9bcb8e3ffbc6a6f7389bb93ffc920577d"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "fdeflate"
+version = "0.3.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1e6853b52649d4ac5c0bd02320cddc5ba956bdb407c4b75a2c6b75bf51500f8c"
+dependencies = [
+ "simd-adler32",
+]
+
 [[package]]
 name = "filetime"
 version = "0.2.26"
@@ -1360,6 +1560,16 @@ dependencies = [
  "syn",
 ]
 
+[[package]]
+name = "gif"
+version = "0.14.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f5df2ba84018d80c213569363bdcd0c64e6933c67fe4c1d60ecf822971a3c35e"
+dependencies = [
+ "color_quant",
+ "weezl",
+]
+
 [[package]]
 name = "gimli"
 version = "0.31.1"
@@ -1717,6 +1927,46 @@ dependencies = [
  "icu_properties",
 ]
 
+[[package]]
+name = "image"
+version = "0.25.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "85ab80394333c02fe689eaf900ab500fbd0c2213da414687ebf995a65d5a6104"
+dependencies = [
+ "bytemuck",
+ "byteorder-lite",
+ "color_quant",
+ "exr",
+ "gif",
+ "image-webp",
+ "moxcms",
+ "num-traits",
+ "png",
+ "qoi",
+ "ravif",
+ "rayon",
+ "rgb",
+ "tiff",
+ "zune-core",
+ "zune-jpeg",
+]
+
+[[package]]
+name = "image-webp"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "525e9ff3e1a4be2fbea1fdf0e98686a6d98b4d8f937e1bf7402245af1909e8c3"
+dependencies = [
+ "byteorder-lite",
+ "quick-error",
+]
+
+[[package]]
+name = "imgref"
+version = "1.12.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e7c5cedc30da3a610cac6b4ba17597bdf7152cf974e8aab3afb3d54455e371c8"
+
 [[package]]
 name = "indexmap"
 version = "2.11.1"
@@ -1762,6 +2012,17 @@ dependencies = [
  "walkdir",
 ]
 
+[[package]]
+name = "interpolate_name"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c34819042dc3d3971c46c2190835914dfbe0c3c13f61449b2997f4e9722dfa60"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "io-uring"
 version = "0.7.10"
@@ -1854,12 +2115,28 @@ version = "1.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "830d08ce1d1d941e6b30645f1a0eb5643013d835ce3779a5fc208261dbe10f55"
 
+[[package]]
+name = "lebe"
+version = "0.5.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7a79a3332a6609480d7d0c9eab957bca6b455b91bb84e66d19f5ff66294b85b8"
+
 [[package]]
 name = "libc"
 version = "0.2.175"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6a82ae493e598baaea5209805c49bbf2ea7de956d50d7da0da1164f9c6d28543"
 
+[[package]]
+name = "libfuzzer-sys"
+version = "0.4.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f12a681b7dd8ce12bff52488013ba614b869148d54dd79836ab85aafdd53f08d"
+dependencies = [
+ "arbitrary",
+ "cc",
+]
+
 [[package]]
 name = "libloading"
 version = "0.8.8"
@@ -1915,6 +2192,15 @@ version = "0.4.28"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "34080505efa8e45a4b816c349525ebe327ceaa8559756f0356cba97ef3bf7432"
 
+[[package]]
+name = "loop9"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0fae87c125b03c1d2c0150c90365d7d6bcc53fb73a9acaef207d2d065860f062"
+dependencies = [
+ "imgref",
+]
+
 [[package]]
 name = "lru-slab"
 version = "0.1.2"
@@ -1969,6 +2255,16 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "maybe-rayon"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8ea1f30cedd69f0a2954655f7188c6a834246d2bcf1e315e2ac40c4b24dc9519"
+dependencies = [
+ "cfg-if",
+ "rayon",
+]
+
 [[package]]
 name = "memchr"
 version = "2.7.5"
@@ -2019,6 +2315,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1fa76a2c86f704bdb222d66965fb3d63269ce38518b83cb0575fca855ebb6316"
 dependencies = [
  "adler2",
+ "simd-adler32",
 ]
 
 [[package]]
@@ -2053,6 +2350,16 @@ dependencies = [
  "syn",
 ]
 
+[[package]]
+name = "moxcms"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bb85c154ba489f01b25c0d36ae69a87e4a1c73a72631fc6c0eb6dde34a73e44b"
+dependencies = [
+ "num-traits",
+ "pxfm",
+]
+
 [[package]]
 name = "native-tls"
 version = "0.2.14"
@@ -2070,6 +2377,12 @@ dependencies = [
  "tempfile",
 ]
 
+[[package]]
+name = "new_debug_unreachable"
+version = "1.0.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "650eef8c711430f1a879fdd01d4745a7deea475becfb90269c06775983bbf086"
+
 [[package]]
 name = "nom"
 version = "7.1.3"
@@ -2080,6 +2393,21 @@ dependencies = [
  "minimal-lexical",
 ]
 
+[[package]]
+name = "nom"
+version = "8.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "df9761775871bdef83bee530e60050f7e54b1105350d6884eb0fb4f46c2f9405"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "noop_proc_macro"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0676bb32a98c1a483ce53e500a81ad9c3d5b3f7c920c28c24e9cb0980d0b5bc8"
+
 [[package]]
 name = "num"
 version = "0.4.3"
@@ -2114,6 +2442,17 @@ dependencies = [
  "num-traits",
 ]
 
+[[package]]
+name = "num-derive"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ed3955f1a9c7c0c15e092f9c887db08b1fc683305fdf6eb6684f22555355e202"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "num-integer"
 version = "0.1.46"
@@ -2386,6 +2725,12 @@ version = "1.0.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "57c0d7b74b563b49d38dae00a0c37d4d6de9b432382b2892f0574ddcae73fd0a"
 
+[[package]]
+name = "pastey"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "35fb2e5f958ec131621fdd531e9fc186ed768cbe395337403ae56c17a74c68ec"
+
 [[package]]
 name = "percent-encoding"
 version = "2.3.2"
@@ -2410,6 +2755,19 @@ version = "0.3.32"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7edddbd0b52d732b21ad9a5fab5c704c14cd949e5e9a1ec5929a24fded1b904c"
 
+[[package]]
+name = "png"
+version = "0.18.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "60769b8b31b2a9f263dae2776c37b1b28ae246943cf719eb6946a1db05128a61"
+dependencies = [
+ "bitflags 2.9.4",
+ "crc32fast",
+ "fdeflate",
+ "flate2",
+ "miniz_oxide",
+]
+
 [[package]]
 name = "portable-atomic"
 version = "1.11.1"
@@ -2475,6 +2833,25 @@ dependencies = [
  "unicode-ident",
 ]
 
+[[package]]
+name = "profiling"
+version = "1.0.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3eb8486b569e12e2c32ad3e204dbaba5e4b5b216e9367044f25f1dba42341773"
+dependencies = [
+ "profiling-procmacros",
+]
+
+[[package]]
+name = "profiling-procmacros"
+version = "1.0.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "52717f9a02b6965224f95ca2a81e2e0c5c43baacd28ca057577988930b6c3d5b"
+dependencies = [
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "pulp"
 version = "0.21.5"
@@ -2512,6 +2889,27 @@ version = "0.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "40e24eee682d89fb193496edf918a7f407d30175b2e785fe057e4392dfd182e0"
 
+[[package]]
+name = "pxfm"
+version = "0.1.28"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b5a041e753da8b807c9255f28de81879c78c876392ff2469cde94799b2896b9d"
+
+[[package]]
+name = "qoi"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f6d64c71eb498fe9eae14ce4ec935c555749aef511cca85b5568910d6e48001"
+dependencies = [
+ "bytemuck",
+]
+
+[[package]]
+name = "quick-error"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a993555f31e5a609f617c12db6250dedcac1b0a85076912c436e6fc9b2c8e6a3"
+
 [[package]]
 name = "quinn"
 version = "0.11.9"
@@ -2651,6 +3049,56 @@ dependencies = [
  "rand 0.9.2",
 ]
 
+[[package]]
+name = "rav1e"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "43b6dd56e85d9483277cde964fd1bdb0428de4fec5ebba7540995639a21cb32b"
+dependencies = [
+ "aligned-vec",
+ "arbitrary",
+ "arg_enum_proc_macro",
+ "arrayvec",
+ "av-scenechange",
+ "av1-grain",
+ "bitstream-io",
+ "built",
+ "cfg-if",
+ "interpolate_name",
+ "itertools 0.14.0",
+ "libc",
+ "libfuzzer-sys",
+ "log",
+ "maybe-rayon",
+ "new_debug_unreachable",
+ "noop_proc_macro",
+ "num-derive",
+ "num-traits",
+ "paste",
+ "profiling",
+ "rand 0.9.2",
+ "rand_chacha 0.9.0",
+ "simd_helpers",
+ "thiserror 2.0.16",
+ "v_frame",
+ "wasm-bindgen",
+]
+
+[[package]]
+name = "ravif"
+version = "0.13.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e52310197d971b0f5be7fe6b57530dcd27beb35c1b013f29d66c1ad73fbbcc45"
+dependencies = [
+ "avif-serialize",
+ "imgref",
+ "loop9",
+ "quick-error",
+ "rav1e",
+ "rayon",
+ "rgb",
+]
+
 [[package]]
 name = "raw-cpuid"
 version = "11.6.0"
@@ -2834,6 +3282,12 @@ dependencies = [
  "webpki-roots 1.0.2",
 ]
 
+[[package]]
+name = "rgb"
+version = "0.8.53"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "47b34b781b31e5d73e9fbc8689c70551fd1ade9a19e3e28cfec8580a79290cc4"
+
 [[package]]
 name = "ring"
 version = "0.17.14"
@@ -3103,6 +3557,21 @@ version = "1.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0fda2ff0d084019ba4d7c6f371c95d8fd75ce3524c3cb8fb653a3023f6323e64"
 
+[[package]]
+name = "simd-adler32"
+version = "0.3.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "703d5c7ef118737c72f1af64ad2f6f8c5e1921f818cdcb97b8fe6fc69bf66214"
+
+[[package]]
+name = "simd_helpers"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "95890f873bec569a0362c235787f3aca6e1e887302ba4840839bcc6459c42da6"
+dependencies = [
+ "quote",
+]
+
 [[package]]
 name = "slab"
 version = "0.4.11"
@@ -3143,7 +3612,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5851699c4033c63636f7ea4cf7b7c1f1bf06d0cc03cfb42e711de5a5c46cf326"
 dependencies = [
  "base64 0.13.1",
- "nom",
+ "nom 7.1.3",
  "serde",
  "unicode-segmentation",
 ]
@@ -3321,6 +3790,20 @@ dependencies = [
  "syn",
 ]
 
+[[package]]
+name = "tiff"
+version = "0.11.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b63feaf3343d35b6ca4d50483f94843803b0f51634937cc2ec519fc32232bc52"
+dependencies = [
+ "fax",
+ "flate2",
+ "half",
+ "quick-error",
+ "weezl",
+ "zune-jpeg",
+]
+
 [[package]]
 name = "tinystr"
 version = "0.8.1"
@@ -3752,6 +4235,17 @@ dependencies = [
  "wasm-bindgen",
 ]
 
+[[package]]
+name = "v_frame"
+version = "0.3.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "666b7727c8875d6ab5db9533418d7c764233ac9c0cff1d469aec8fa127597be2"
+dependencies = [
+ "aligned-vec",
+ "num-traits",
+ "wasm-bindgen",
+]
+
 [[package]]
 name = "vcpkg"
 version = "0.2.15"
@@ -3936,6 +4430,12 @@ dependencies = [
  "rustls-pki-types",
 ]
 
+[[package]]
+name = "weezl"
+version = "0.1.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a28ac98ddc8b9274cb41bb4d9d4d5c425b6020c50c46f25559911905610b4a88"
+
 [[package]]
 name = "winapi"
 version = "0.3.9"
@@ -4323,6 +4823,12 @@ dependencies = [
  "rustix",
 ]
 
+[[package]]
+name = "y4m"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7a5a4b21e1a62b67a2970e6831bc091d7b87e119e7f9791aef9702e3bef04448"
+
 [[package]]
 name = "yoke"
 version = "0.7.5"
@@ -4461,3 +4967,27 @@ dependencies = [
  "memchr",
  "typed-path",
 ]
+
+[[package]]
+name = "zune-core"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cb8a0807f7c01457d0379ba880ba6322660448ddebc890ce29bb64da71fb40f9"
+
+[[package]]
+name = "zune-inflate"
+version = "0.2.54"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "73ab332fe2f6680068f3582b16a24f90ad7096d5d39b974d1c0aff0125116f02"
+dependencies = [
+ "simd-adler32",
+]
+
+[[package]]
+name = "zune-jpeg"
+version = "0.5.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "27bc9d5b815bc103f142aa054f561d9187d191692ec7c2d1e2b4737f8dbd7296"
+dependencies = [
+ "zune-core",
+]

data/README.md

@@ -942,6 +942,49 @@ All NER methods return entities in a consistent format:
 }
 ```
 
+## Vision-Language Models (VLM)
+
+Red-Candle supports vision-language models for understanding and describing images. The VLM module uses LLaVA (Large Language and Vision Assistant), which combines a CLIP vision encoder with a Llama language model.
+
+### Basic Usage
+
+```ruby
+require 'candle'
+
+# Load a LLaVA model (requires ~13GB download on first use)
+vlm = Candle::VLM.from_pretrained("llava-hf/llava-v1.6-vicuna-7b-hf")
+
+# Describe an image
+description = vlm.describe("photo.jpg")
+
+# Ask a question about an image
+answer = vlm.ask("photo.jpg", "What animal is in this image?")
+# => "The animal in the image is a cat."
+
+# Control output length
+vlm.describe("photo.jpg", max_length: 500)
+vlm.ask("photo.jpg", "What colors do you see?", max_length: 50)
+```
+
+### How It Works
+
+1. **CLIP Vision Encoder**: Converts the image into a sequence of visual feature tokens (576 patches from a 336x336 image)
+2. **MM Projector**: Projects vision features into the language model's embedding space
+3. **Llama LLM**: Processes the combined image+text embeddings and generates a text response
+
+### Supported Models
+
+| Model | LLM Backend | Size | Notes |
+|:------|:-----------|:-----|:------|
+| `llava-hf/llava-v1.6-vicuna-7b-hf` | Llama (Vicuna) | 13GB | Recommended, LLaVA-Next with Llama backend |
+
+### Notes
+
+- First load downloads ~13GB of model weights (cached for subsequent use)
+- Image preprocessing is automatic (resize, normalize to CLIP format)
+- Generation uses greedy decoding
+- Multiple calls work correctly (KV cache is reset between queries)
+
 ## Common Runtime Errors
 
 ### Weight is negative, too large or not a valid number

data/ext/candle/Cargo.toml

@@ -22,6 +22,7 @@ serde = { version = "1.0", features = ["derive"] }
 tokio = { version = "1.45", features = ["rt", "macros"] }
 rand = "0.8"
 outlines-core = "0.2.11"
+image = "0.25"
 
 [features]
 default = []

data/ext/candle/src/lib.rs

@@ -46,6 +46,7 @@ fn init(ruby: &Ruby) -> Result<()> {
     ruby::init_llm(rb_candle)?;
     ruby::ner::init(rb_candle)?;
     ruby::reranker::init(rb_candle)?;
+    ruby::vlm::init(rb_candle)?;
     ruby::dtype::init(rb_candle)?;
     ruby::device::init(rb_candle)?;
     ruby::tensor::init(rb_candle)?;

data/ext/candle/src/ruby/mod.rs

@@ -10,6 +10,7 @@ pub mod tokenizer;
 pub mod structured;
 pub mod reranker;
 pub mod ner;
+pub mod vlm;
 
 pub use embedding_model::{EmbeddingModel, EmbeddingModelInner};
 pub use tensor::Tensor;

data/ext/candle/src/ruby/vlm.rs

@@ -0,0 +1,330 @@
+use magnus::{function, method, prelude::*, Error, RModule, Ruby};
+use candle_transformers::models::llava::{
+    config::{LLaVAConfig, HFLLaVAConfig, HFGenerationConfig, HFPreProcessorConfig},
+    LLaVA,
+};
+use candle_transformers::models::llama::Cache;
+use candle_core::{Device as CoreDevice, Tensor, DType};
+use candle_nn::VarBuilder;
+use hf_hub::{api::sync::Api, Repo, RepoType};
+use tokenizers::Tokenizer;
+use crate::ruby::{Device, Result};
+use crate::tokenizer::TokenizerWrapper;
+
+const CLIP_MEAN: [f32; 3] = [0.48145466, 0.4578275, 0.40821073];
+const CLIP_STD: [f32; 3] = [0.26862954, 0.26130258, 0.27577711];
+
+/// Vision-Language Model wrapping LLaVA for image understanding.
+/// Uses CLIP vision encoder + MM projector + Llama LLM.
+///
+/// Note: LLaVA contains trait objects (dyn Module) that are !Send,
+/// so we wrap it in an UnsafeCell. This is safe because Ruby's GVL
+/// ensures single-threaded access to the model.
+struct UnsafeSendSync<T>(T);
+unsafe impl<T> Send for UnsafeSendSync<T> {}
+unsafe impl<T> Sync for UnsafeSendSync<T> {}
+
+#[magnus::wrap(class = "Candle::VLM", free_immediately, size)]
+pub struct VLM {
+    model: std::cell::RefCell<UnsafeSendSync<LLaVA>>,
+    tokenizer: TokenizerWrapper,
+    cache: std::cell::RefCell<UnsafeSendSync<Cache>>,
+    config: LLaVAConfig,
+    device: CoreDevice,
+    model_id: String,
+    image_size: usize,
+    eos_token_id: u32,
+}
+
+impl VLM {
+    pub fn new(model_id: String, device: Option<Device>) -> Result<Self> {
+        let device = device.unwrap_or(Device::best()).as_device()?;
+        Self::load_model(model_id, device)
+    }
+
+    fn load_model(model_id: String, device: CoreDevice) -> std::result::Result<Self, Error> {
+        let ruby = Ruby::get().unwrap();
+        let runtime_error = ruby.exception_runtime_error();
+
+        let result = (|| -> std::result::Result<_, Box<dyn std::error::Error + Send + Sync>> {
+            let api = Api::new()?;
+            let repo = api.repo(Repo::new(model_id.clone(), RepoType::Model));
+
+            // Download config files
+            let config_filename = repo.get("config.json")?;
+            let gen_config_filename = repo.get("generation_config.json")?;
+            let preproc_config_filename = repo.get("preprocessor_config.json")?;
+            let tokenizer_filename = repo.get("tokenizer.json")?;
+
+            // Read configs
+            let config_str = std::fs::read_to_string(&config_filename)?;
+            let gen_config_str = std::fs::read_to_string(&gen_config_filename)?;
+            let preproc_config_str = std::fs::read_to_string(&preproc_config_filename)?;
+
+            // Patch config: some models have null pad_token_id in text_config
+            // but candle's HFLLaVATextConfig requires usize. Fix by defaulting to 0.
+            let mut config_json: serde_json::Value = serde_json::from_str(&config_str)?;
+            let top_pad_id = config_json.get("pad_token_id")
+                .and_then(|v| v.as_u64())
+                .unwrap_or(0);
+            // Patch missing image_grid_pinpoints for LLaVA 1.5
+            if config_json.get("image_grid_pinpoints").map_or(true, |v| v.is_null()) {
+                config_json["image_grid_pinpoints"] = serde_json::json!([[336, 672], [672, 336], [672, 672]]);
+            }
+            if let Some(text_config) = config_json.get_mut("text_config") {
+                if text_config.get("pad_token_id").map_or(true, |v| v.is_null()) {
+                    text_config["pad_token_id"] = serde_json::Value::Number(top_pad_id.into());
+                }
+            }
+            let patched_config_str = serde_json::to_string(&config_json)?;
+            let hf_config: HFLLaVAConfig = serde_json::from_str(&patched_config_str)?;
+            let gen_config: HFGenerationConfig = serde_json::from_str(&gen_config_str)?;
+            let preproc_config: HFPreProcessorConfig = serde_json::from_str(&preproc_config_str)?;
+
+            let image_size = hf_config.vision_config.image_size;
+            let eos_token_id = gen_config.eos_token_id as u32;
+
+            let clip_vision_config = hf_config.to_clip_vision_config();
+            let config = hf_config.to_llava_config(&gen_config, &preproc_config);
+
+            // Load tokenizer
+            let tokenizer = Tokenizer::from_file(tokenizer_filename)?;
+
+            // Download weight files (sharded)
+            let weight_files = Self::download_weights(&repo)?;
+
+            // Load model weights
+            let vb = unsafe {
+                VarBuilder::from_mmaped_safetensors(&weight_files, DType::F32, &device)?
+            };
+
+            // Load LLaVA model with CLIP vision config
+            let model = LLaVA::load(vb, &config, Some(clip_vision_config))?;
+
+            // Create KV cache for the Llama LLM
+            let llama_config = config.to_llama_config();
+            let cache = Cache::new(true, DType::F32, &llama_config, &device)?;
+
+            Ok((model, TokenizerWrapper::new(tokenizer), cache, config, image_size, eos_token_id))
+        })();
+
+        match result {
+            Ok((model, tokenizer, cache, config, image_size, eos_token_id)) => {
+                Ok(Self {
+                    model: std::cell::RefCell::new(UnsafeSendSync(model)),
+                    tokenizer,
+                    cache: std::cell::RefCell::new(UnsafeSendSync(cache)),
+                    config,
+                    device,
+                    model_id,
+                    image_size,
+                    eos_token_id,
+                })
+            }
+            Err(e) => Err(Error::new(runtime_error, format!("Failed to load VLM: {}", e))),
+        }
+    }
+
+    fn download_weights(
+        repo: &hf_hub::api::sync::ApiRepo,
+    ) -> std::result::Result<Vec<std::path::PathBuf>, Box<dyn std::error::Error + Send + Sync>> {
+        // Try single file first
+        if let Ok(path) = repo.get("model.safetensors") {
+            return Ok(vec![path]);
+        }
+
+        // Try to get the index file for sharded weights
+        let index_path = repo.get("model.safetensors.index.json")?;
+        let index_str = std::fs::read_to_string(&index_path)?;
+        let index: serde_json::Value = serde_json::from_str(&index_str)?;
+
+        let weight_map = index["weight_map"].as_object()
+            .ok_or("Missing weight_map in index")?;
+
+        let mut filenames: Vec<String> = weight_map.values()
+            .filter_map(|v| v.as_str().map(String::from))
+            .collect();
+        filenames.sort();
+        filenames.dedup();
+
+        let mut paths = Vec::new();
+        for filename in &filenames {
+            let path = repo.get(filename)?;
+            paths.push(path);
+        }
+
+        Ok(paths)
+    }
+
+    /// Load and preprocess an image from a file path into a CLIP-ready tensor
+    fn load_image(&self, image_path: &str) -> std::result::Result<Tensor, Error> {
+        let ruby = Ruby::get().unwrap();
+        let runtime_error = ruby.exception_runtime_error();
+
+        let img = image::open(image_path)
+            .map_err(|e| Error::new(runtime_error, format!("Failed to open image: {}", e)))?;
+
+        // Resize to expected size
+        let img = img.resize_exact(
+            self.image_size as u32,
+            self.image_size as u32,
+            image::imageops::FilterType::Triangle,
+        );
+
+        let img = img.to_rgb8();
+        let (width, height) = img.dimensions();
+        let h = height as usize;
+        let w = width as usize;
+
+        // Convert to CHW format with CLIP normalization
+        let mut chw = vec![0f32; 3 * h * w];
+        for y in 0..h {
+            for x in 0..w {
+                let p = img.get_pixel(x as u32, y as u32);
+                chw[0 * h * w + y * w + x] = (p[0] as f32 / 255.0 - CLIP_MEAN[0]) / CLIP_STD[0];
+                chw[1 * h * w + y * w + x] = (p[1] as f32 / 255.0 - CLIP_MEAN[1]) / CLIP_STD[1];
+                chw[2 * h * w + y * w + x] = (p[2] as f32 / 255.0 - CLIP_MEAN[2]) / CLIP_STD[2];
+            }
+        }
+
+        Tensor::from_vec(chw, (1, 3, h, w), &self.device)
+            .map_err(|e| Error::new(runtime_error, format!("Failed to create image tensor: {}", e)))
+    }
+
+    /// Describe an image
+    pub fn describe(&self, image_path: String, max_length: Option<usize>) -> std::result::Result<String, Error> {
+        self.ask(image_path, "Describe this image in detail.".to_string(), max_length)
+    }
+
+    /// Ask a question about an image
+    pub fn ask(&self, image_path: String, question: String, max_length: Option<usize>) -> std::result::Result<String, Error> {
+        let ruby = Ruby::get().unwrap();
+        let runtime_error = ruby.exception_runtime_error();
+        let max_length = max_length.unwrap_or(256);
+
+        // Load and preprocess image
+        let image_tensor = self.load_image(&image_path)?;
+
+        // Build prompt with image token placeholder
+        // LLaVA 1.5 HF format: USER: <image>\n{question}\nASSISTANT:
+        let prompt = format!("USER: <image>\n{}\nASSISTANT:", question);
+
+        // Tokenize
+        let encoding = self.tokenizer.inner().encode(prompt.as_str(), false)
+            .map_err(|e| Error::new(runtime_error, format!("Tokenization failed: {}", e)))?;
+        let input_ids: Vec<u32> = encoding.get_ids().to_vec();
+
+        // LLaVA expects I64 input IDs
+        let input_ids_i64: Vec<i64> = input_ids.iter().map(|&id| id as i64).collect();
+        let input_tensor = Tensor::new(&input_ids_i64[..], &self.device)
+            .map_err(|e| Error::new(runtime_error, format!("Failed to create input tensor: {}", e)))?
+            .unsqueeze(0)
+            .map_err(|e| Error::new(runtime_error, format!("Failed to unsqueeze: {}", e)))?;
+
+        let mut model_ref = self.model.borrow_mut();
+        let model = &mut model_ref.0;
+        let mut cache_ref = self.cache.borrow_mut();
+        let cache = &mut cache_ref.0;
+
+        // Prepare multimodal input: merge image features with text embeddings
+        let image_size = (self.image_size as u32, self.image_size as u32);
+        let input_embeds = model.prepare_inputs_labels_for_multimodal(
+            &input_tensor,
+            &[image_tensor],
+            &[image_size],
+        ).map_err(|e| Error::new(runtime_error, format!("Failed to prepare multimodal input: {}", e)))?;
+
+        // Reset KV cache for fresh generation
+        let llama_config = self.config.to_llama_config();
+        *cache = Cache::new(true, DType::F32, &llama_config, &self.device)
+            .map_err(|e| Error::new(runtime_error, format!("Failed to reset cache: {}", e)))?;
+
+        // Generate tokens autoregressively
+        let mut generated_tokens: Vec<u32> = Vec::new();
+        let mut current_embeds = input_embeds;
+        let mut pos = 0usize;
+
+        for _i in 0..max_length {
+            let logits = model.forward(&current_embeds, pos, cache)
+                .map_err(|e| Error::new(runtime_error, format!("Forward pass failed at pos {}: {}", pos, e)))?;
+
+            // Advance position by the number of tokens we just processed
+            let step_len = current_embeds.dim(1)
+                .map_err(|e| Error::new(runtime_error, format!("Failed to get step len: {}", e)))?;
+            pos += step_len;
+
+            // Get logits for last position
+            let logits = logits.flatten_all()
+                .map_err(|e| Error::new(runtime_error, format!("Failed to flatten: {}", e)))?;
+
+            // Handle multi-dim logits (take last token if needed)
+            let vocab_size = self.config.vocab_size;
+            let logits = if logits.elem_count() > vocab_size {
+                let n_tokens = logits.elem_count() / vocab_size;
+                logits.reshape((n_tokens, vocab_size))
+                    .map_err(|e| Error::new(runtime_error, format!("Failed to reshape logits: {}", e)))?
+                    .narrow(0, n_tokens - 1, 1)
+                    .map_err(|e| Error::new(runtime_error, format!("Failed to narrow logits: {}", e)))?
+                    .squeeze(0)
+                    .map_err(|e| Error::new(runtime_error, format!("Failed to squeeze logits: {}", e)))?
+            } else {
+                logits
+            };
+
+            let logits = logits.to_dtype(DType::F32)
+                .map_err(|e| Error::new(runtime_error, format!("Failed to convert dtype: {}", e)))?;
+
+            // Greedy decoding
+            let next_token = logits.argmax(0)
+                .map_err(|e| Error::new(runtime_error, format!("Argmax failed: {}", e)))?
+                .to_scalar::<u32>()
+                .map_err(|e| Error::new(runtime_error, format!("Failed to get token: {}", e)))?;
+
+            // Check for EOS
+            if next_token == self.eos_token_id {
+                break;
+            }
+
+            generated_tokens.push(next_token);
+
+            // For subsequent tokens, embed directly through Llama's embedding layer
+            let next_input = Tensor::new(&[next_token as i64], &self.device)
+                .map_err(|e| Error::new(runtime_error, format!("Failed to create next input: {}", e)))?
+                .unsqueeze(0)
+                .map_err(|e| Error::new(runtime_error, format!("Failed to unsqueeze next: {}", e)))?;
+
+            current_embeds = model.llama.embed(&next_input)
+                .map_err(|e| Error::new(runtime_error, format!("Failed to embed next token: {}", e)))?;
+        }
+
+        // Decode generated tokens
+        let text = self.tokenizer.inner().decode(&generated_tokens, true)
+            .map_err(|e| Error::new(runtime_error, format!("Decoding failed: {}", e)))?;
+
+        Ok(text.trim().to_string())
+    }
+
+    pub fn model_id(&self) -> String {
+        self.model_id.clone()
+    }
+
+    pub fn device(&self) -> Device {
+        Device::from_device(&self.device)
+    }
+
+    pub fn tokenizer(&self) -> std::result::Result<crate::ruby::tokenizer::Tokenizer, Error> {
+        Ok(crate::ruby::tokenizer::Tokenizer(self.tokenizer.clone()))
+    }
+}
+
+pub fn init(rb_candle: RModule) -> std::result::Result<(), Error> {
+    let ruby = Ruby::get().unwrap();
+    let c_vlm = rb_candle.define_class("VLM", ruby.class_object())?;
+    c_vlm.define_singleton_method("_create", function!(VLM::new, 2))?;
+    c_vlm.define_method("_describe", method!(VLM::describe, 2))?;
+    c_vlm.define_method("_ask", method!(VLM::ask, 3))?;
+    c_vlm.define_method("model_id", method!(VLM::model_id, 0))?;
+    c_vlm.define_method("device", method!(VLM::device, 0))?;
+    c_vlm.define_method("tokenizer", method!(VLM::tokenizer, 0))?;
+    Ok(())
+}

data/lib/candle/version.rb

@@ -1,5 +1,5 @@
 # :nocov:
 module Candle
-  VERSION = "1.6.1"
+  VERSION = "1.7.0"
 end
 # :nocov:

data/lib/candle/vlm.rb

@@ -0,0 +1,31 @@
+# frozen_string_literal: true
+
+module Candle
+  class VLM
+    class << self
+      def from_pretrained(model_id, device: nil, **options)
+        device_obj = case device
+                     when "cpu" then Candle::Device.cpu
+                     when "metal" then Candle::Device.metal
+                     when "cuda" then Candle::Device.cuda
+                     when Candle::Device then device
+                     when nil then nil
+                     else Candle::Device.best
+                     end
+        _create(model_id, device_obj)
+      end
+    end
+
+    def describe(image_path, max_length: 256)
+      _describe(image_path, max_length)
+    end
+
+    def ask(image_path, question, max_length: 256)
+      _ask(image_path, question, max_length)
+    end
+
+    def inspect
+      "#<Candle::VLM model_id=#{model_id.inspect} device=#{device}>"
+    end
+  end
+end

data/lib/candle.rb

@@ -11,4 +11,5 @@ require_relative "candle/llm"
 require_relative "candle/agent"
 require_relative "candle/tokenizer"
 require_relative "candle/ner"
+require_relative "candle/vlm"
 require_relative "candle/build_info"

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: red-candle
 version: !ruby/object:Gem::Version
-  version: 1.6.1
+  version: 1.7.0
 platform: ruby
 authors:
 - Christopher Petersen
@@ -243,6 +243,7 @@ files:
 - ext/candle/src/ruby/tensor.rs
 - ext/candle/src/ruby/tokenizer.rs
 - ext/candle/src/ruby/utils.rs
+- ext/candle/src/ruby/vlm.rs
 - ext/candle/src/structured/integration_test.rs
 - ext/candle/src/structured/mod.rs
 - ext/candle/src/structured/schema_processor.rs
@@ -268,6 +269,7 @@ files:
 - lib/candle/tool.rb
 - lib/candle/tool_call_parser.rb
 - lib/candle/version.rb
+- lib/candle/vlm.rb
 - lib/red-candle.rb
 homepage: https://github.com/scientist-labs/red-candle
 licenses: