cuttlefish_abi/lib.rs
1//! The contract between the cuttlefish host and its guest proc-blocks.
2//!
3//! Both sides depend on this crate precisely so that they cannot drift: a block
4//! is compiled separately from the host, often at a different time by a
5//! different person, and the only thing keeping them able to talk is that they
6//! agreed on these types.
7//!
8//! # Why a command loop, not function calls
9//!
10//! A block does not call the host. It *returns* a [`Command`] describing what it
11//! wants done, and the host — after doing it — hands back an [`Event`] and asks
12//! for the next command. Control is inverted relative to the obvious design, and
13//! not for taste:
14//!
15//! - A core-wasm guest is single-threaded and offers no execution context the
16//! host could call back into while the guest is blocked. A "call the host and
17//! wait" design has nowhere to deliver the answer.
18//! - Inference must run on a different thread from the wasm store, which is
19//! `!Sync` and cannot be touched from there.
20//! - Because the host decides whether to take the next step, cancellation needs
21//! no cooperation from the guest at all: the host simply stops stepping. A
22//! guest cannot ignore, delay, or trap its way out of being cancelled.
23//!
24//! Everything crosses the boundary as JSON. That is slower than a packed binary
25//! layout, deliberately: the boundary stays inspectable, a mismatch produces a
26//! legible error rather than a misread integer, and the volume is low because
27//! bulk data does not cross it. Revisit only if profiling says to.
28//!
29//! # Why bulk data does not cross this boundary
30//!
31//! No command hands a block the contents of a file. A block [`Command::Open`]s a
32//! path, receives a [`Handle`] and a length, then pulls bounded windows with
33//! [`Command::Slice`].
34//!
35//! This keeps guest memory proportional to the window a block chooses rather
36//! than to the size of its input. A block written against a small file behaves
37//! identically against a huge one, and the 4 GiB ceiling of 32-bit wasm stops
38//! being something block authors must reason about — which is what lets this
39//! project stay on `wasm32` instead of paying for `wasm64`.
40
41#![forbid(unsafe_code)]
42#![warn(missing_docs)]
43
44use serde::{Deserialize, Serialize};
45
46/// A job-scoped reference to something the host holds open for a guest.
47///
48/// Job-scoping is a security property, not bookkeeping. A handle table lives and
49/// dies with a single job, so a handle from one job names nothing in another.
50/// That is why [`Command::Slice`] carries no path and needs no capability check
51/// of its own: the check happened once, at [`Command::Open`], and a handle
52/// cannot be forged into a reference to another job's data.
53pub type Handle = u32;
54
55/// The shape of a value flowing through a pipeline.
56///
57/// Deliberately small. This exists to catch the mistake that actually happens
58/// when blocks are composed — one block emitting a summary string into another
59/// expecting a list of chunks — not to be a general-purpose type system. A
60/// richer one would need inference, and inference over a language with no
61/// expressions is machinery without a use.
62/// Written and read as a compact string — `text`, `[text]`, `{path: text}` —
63/// rather than as a nested tagged object.
64///
65/// Two reasons, and the second is the one that forced it. It reads well in an
66/// error message and in a spec, so one syntax serves the wire, the diagnostics,
67/// and the DSL. And a recursive enum serialized structurally makes serde's
68/// generic serializer recurse deeply enough to blow rustc's recursion limit in
69/// the *guest* crate — which would have meant every block author adding
70/// `#![recursion_limit]` to work around a detail of this type.
71#[derive(Debug, Clone, PartialEq, Eq)]
72pub enum Ty {
73 /// A UTF-8 string.
74 Text,
75 /// A JSON number, integral or fractional.
76 ///
77 /// One variant rather than an int/float pair: JSON itself does not
78 /// distinguish them, so a block returning `3` where `3.0` was meant would
79 /// fail a check that exists only in the type system and nowhere in the
80 /// data. Where the distinction matters downstream — a Parquet column type,
81 /// say — it is decided by looking at the values, which is the only place
82 /// the information actually exists.
83 Number,
84 /// A JSON boolean.
85 Bool,
86 /// Opaque bytes, base64-encoded on the wire.
87 Bytes,
88 /// A handle naming an image the host holds.
89 Image,
90 /// A handle naming a paged document.
91 Document,
92 /// Any JSON value. The top type: everything is assignable to it.
93 ///
94 /// An escape hatch, and one worth using sparingly — a pipeline of `Json`
95 /// seams typechecks unconditionally, which is the same as not checking.
96 Json,
97 /// An ordered sequence.
98 List(Box<Ty>),
99 /// A fixed set of named fields.
100 ///
101 /// A `BTreeMap` so that two records written in different field orders are
102 /// the same type, and so error messages list fields the same way twice.
103 Record(std::collections::BTreeMap<String, Ty>),
104}
105
106impl Ty {
107 /// Whether a value of this type can be fed where `expected` is required.
108 ///
109 /// Not equality: [`Ty::Json`] accepts anything, and a record with *extra*
110 /// fields satisfies one that needs fewer. Both directions of that matter —
111 /// a block that adds a field should not break its consumer, and a block
112 /// that requires a field its producer never emits should fail loudly.
113 pub fn assignable_to(&self, expected: &Ty) -> bool {
114 match (self, expected) {
115 (_, Ty::Json) => true,
116 (Ty::List(a), Ty::List(b)) => a.assignable_to(b),
117 (Ty::Record(have), Ty::Record(need)) => need
118 .iter()
119 .all(|(name, want)| have.get(name).is_some_and(|got| got.assignable_to(want))),
120 (a, b) => a == b,
121 }
122 }
123
124 /// Whether a live JSON value could plausibly be an instance of this
125 /// type — a runtime counterpart to [`Self::assignable_to`], which only
126 /// ever compares two declared `Ty`s against each other. Nothing in the
127 /// host checked a block's *actual* output against what it declared
128 /// until this existed: a block could claim `{summary: text}` and return
129 /// `{text: "..."}` and nothing downstream would notice until whatever
130 /// consumed `summary` got `null`.
131 ///
132 /// Deliberately permissive, not a full validator: [`Ty::Bytes`],
133 /// [`Ty::Image`], and [`Ty::Document`] have no fixed JSON shape defined
134 /// anywhere in this protocol (unlike [`Ty::Text`]/[`Ty::List`]/
135 /// [`Ty::Record`], which map onto JSON strings/arrays/objects
136 /// unambiguously) — inventing a shape for them here risks rejecting
137 /// legitimate values a real block already produces. Those three, like
138 /// [`Ty::Json`], accept anything.
139 pub fn matches_value(&self, value: &serde_json::Value) -> bool {
140 match self {
141 Ty::Json | Ty::Bytes | Ty::Image | Ty::Document => true,
142 Ty::Text => value.is_string(),
143 Ty::Number => value.is_number(),
144 Ty::Bool => value.is_boolean(),
145 Ty::List(inner) => value
146 .as_array()
147 .is_some_and(|items| items.iter().all(|v| inner.matches_value(v))),
148 Ty::Record(fields) => value.as_object().is_some_and(|obj| {
149 fields
150 .iter()
151 .all(|(name, want)| obj.get(name).is_some_and(|v| want.matches_value(v)))
152 }),
153 }
154 }
155
156 /// A short human-readable rendering, for error messages.
157 pub fn describe(&self) -> String {
158 match self {
159 Ty::Text => "text".into(),
160 Ty::Number => "number".into(),
161 Ty::Bool => "bool".into(),
162 Ty::Bytes => "bytes".into(),
163 Ty::Image => "image".into(),
164 Ty::Document => "document".into(),
165 Ty::Json => "json".into(),
166 Ty::List(inner) => format!("[{}]", inner.describe()),
167 Ty::Record(fields) => {
168 let body = fields
169 .iter()
170 .map(|(k, v)| format!("{k}: {}", v.describe()))
171 .collect::<Vec<_>>()
172 .join(", ");
173 format!("{{{body}}}")
174 }
175 }
176 }
177}
178
179impl std::fmt::Display for Ty {
180 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
181 f.write_str(&self.describe())
182 }
183}
184
185impl std::str::FromStr for Ty {
186 type Err = String;
187
188 fn from_str(s: &str) -> Result<Self, Self::Err> {
189 parse_ty(s.trim())
190 }
191}
192
193impl Serialize for Ty {
194 fn serialize<S: serde::Serializer>(&self, s: S) -> Result<S::Ok, S::Error> {
195 s.serialize_str(&self.describe())
196 }
197}
198
199impl<'de> Deserialize<'de> for Ty {
200 fn deserialize<D: serde::Deserializer<'de>>(d: D) -> Result<Self, D::Error> {
201 let raw = String::deserialize(d)?;
202 raw.parse().map_err(serde::de::Error::custom)
203 }
204}
205
206/// Parse the type syntax [`Ty::describe`] produces.
207fn parse_ty(s: &str) -> Result<Ty, String> {
208 let s = s.trim();
209 match s {
210 "text" => return Ok(Ty::Text),
211 "number" => return Ok(Ty::Number),
212 "bool" => return Ok(Ty::Bool),
213 "bytes" => return Ok(Ty::Bytes),
214 "image" => return Ok(Ty::Image),
215 "document" => return Ok(Ty::Document),
216 "json" => return Ok(Ty::Json),
217 _ => {}
218 }
219
220 if let Some(inner) = s.strip_prefix('[').and_then(|r| r.strip_suffix(']')) {
221 return Ok(Ty::List(Box::new(parse_ty(inner)?)));
222 }
223
224 if let Some(body) = s.strip_prefix('{').and_then(|r| r.strip_suffix('}')) {
225 let mut fields = std::collections::BTreeMap::new();
226 if !body.trim().is_empty() {
227 for part in split_fields(body) {
228 let (name, ty) = part
229 .split_once(':')
230 .ok_or_else(|| format!("expected `name: type` in `{part}`"))?;
231 fields.insert(name.trim().to_string(), parse_ty(ty)?);
232 }
233 }
234 return Ok(Ty::Record(fields));
235 }
236
237 Err(format!("`{s}` is not a type"))
238}
239
240/// Split record fields on commas that are not inside a nested `[]` or `{}`.
241///
242/// A plain `split(',')` would cut `{a: [x, y]}` in the wrong place.
243fn split_fields(body: &str) -> Vec<String> {
244 let (mut out, mut depth, mut current) = (Vec::new(), 0i32, String::new());
245 for c in body.chars() {
246 match c {
247 '[' | '{' => {
248 depth += 1;
249 current.push(c);
250 }
251 ']' | '}' => {
252 depth -= 1;
253 current.push(c);
254 }
255 ',' if depth == 0 => out.push(std::mem::take(&mut current)),
256 _ => current.push(c),
257 }
258 }
259 if !current.trim().is_empty() {
260 out.push(current);
261 }
262 out
263}
264
265/// What a block accepts and produces.
266///
267/// Declared by the block itself, through a `cf_signature` export, rather than in
268/// a sidecar file beside it. A sidecar can disagree with the code it describes
269/// and nothing forces anyone to notice; a declaration compiled into the module
270/// travels with it, cannot go stale, and leaves one artifact to ship rather than
271/// two to keep in step.
272///
273/// `Display`/`FromStr` render and parse it as `"{input} -> {output}"` —
274/// each side is a [`Ty`], and this is the compact form the catalog caches
275/// and a bundle manifest embeds. Splitting on `" -> "` is unambiguous only
276/// because `Ty::describe()` never produces that substring itself.
277#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
278pub struct Signature {
279 /// What the block needs as input.
280 pub input: Ty,
281 /// What it produces.
282 pub output: Ty,
283}
284
285impl std::fmt::Display for Signature {
286 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
287 write!(f, "{} -> {}", self.input, self.output)
288 }
289}
290
291impl std::str::FromStr for Signature {
292 type Err = String;
293
294 fn from_str(s: &str) -> Result<Self, Self::Err> {
295 let (input, output) = s
296 .split_once(" -> ")
297 .ok_or_else(|| format!("`{s}` is not a signature (expected `input -> output`)"))?;
298 Ok(Signature {
299 input: input.parse()?,
300 output: output.parse()?,
301 })
302 }
303}
304
305/// What kind of thing a handle refers to, reported by [`Event::Opened`].
306///
307/// A block needs this to know which commands are worth issuing: [`Command::Slice`]
308/// on a PNG is a mistake, and [`Command::PageText`] on a plain text file is
309/// meaningless. Reporting it up front means a block can branch on what it
310/// actually got rather than guessing from a file extension.
311#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
312#[serde(tag = "kind", rename_all = "snake_case")]
313pub enum MediaKind {
314 /// Valid UTF-8. Both [`Command::Slice`] and [`Command::SliceBytes`] work.
315 Text,
316 /// An image the host recognised. Usable as an [`Command::Infer`] image.
317 Image {
318 /// Format as detected from content, e.g. `png`, `jpeg`.
319 format: String,
320 },
321 /// A paged document — a PDF, say.
322 Document {
323 /// How many pages it has.
324 pages: u32,
325 /// Whether it carries an extractable text layer.
326 ///
327 /// False for a scanned document, where the only way to read it is to
328 /// rasterize pages and hand them to a vision model. A block that checks
329 /// this can pick the cheap path when it exists and the expensive one
330 /// when it must, instead of silently extracting nothing.
331 has_text_layer: bool,
332 },
333 /// Bytes the host could not classify. Only [`Command::SliceBytes`] applies.
334 Binary,
335}
336
337/// One transformation [`Command::ImageOp`] can apply.
338///
339/// Deliberately a closed set rather than a general filter language: each of
340/// these answers a question an analysis pipeline actually asks, and a closed
341/// set is one a host can implement completely and a block can rely on.
342#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
343#[serde(tag = "op", rename_all = "snake_case")]
344pub enum ImageOperation {
345 /// Scale to fit within these bounds, preserving aspect ratio.
346 ///
347 /// The common case before handing an image to a vision model, which
348 /// resamples to a fixed size anyway — sending a 48-megapixel original
349 /// costs encode time and tokens for detail the model cannot use.
350 Resize {
351 /// Maximum width in pixels.
352 max_width: u32,
353 /// Maximum height in pixels.
354 max_height: u32,
355 },
356 /// Cut out a region, for asking about part of an image rather than all
357 /// of it.
358 Crop {
359 /// Left edge, in pixels from the origin.
360 x: u32,
361 /// Top edge, in pixels from the origin.
362 y: u32,
363 /// Region width.
364 width: u32,
365 /// Region height.
366 height: u32,
367 },
368}
369
370/// What a guest asks the host to do, returned from its `init`/`step` exports.
371#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
372#[serde(tag = "cmd", rename_all = "snake_case")]
373pub enum Command {
374 /// Run a prompt against the job's model.
375 Infer {
376 /// The prompt to generate from.
377 prompt: String,
378 /// Upper bound on tokens generated. A guest can also end generation
379 /// early by returning [`TokenAction::Stop`] from its `on_token` export.
380 max_tokens: u32,
381 /// Images to accompany the prompt, named by handle.
382 ///
383 /// Handles rather than bytes, for the same reason file contents are not
384 /// handed over: an image can be tens of megabytes, and routing it
385 /// through guest memory would put the 4 GiB wasm32 ceiling back in play
386 /// for no benefit. The host already holds the bytes; it can pass them to
387 /// the model directly.
388 ///
389 /// Requires a model with vision capability. Empty for ordinary
390 /// text-only inference, which is why it is `#[serde(default)]` — a block
391 /// compiled before this field existed still deserializes.
392 #[serde(default)]
393 images: Vec<Handle>,
394 },
395 /// Open a file. Capability-checked against the job's spec.
396 ///
397 /// Yields a handle and a length rather than contents — see the crate docs on
398 /// why bulk data does not cross this boundary.
399 Open {
400 /// Path to open. Denied unless the spec grants read access to it.
401 path: String,
402 },
403 /// Pull one bounded window of an open file into guest memory.
404 ///
405 /// The guest picks `len`, so the guest sets its own memory ceiling.
406 Slice {
407 /// Handle from a previous [`Command::Open`].
408 handle: Handle,
409 /// Byte offset to read from. `u64` so that files far larger than a guest
410 /// could hold remain fully addressable.
411 offset: u64,
412 /// Maximum bytes to return. The host may return fewer; see
413 /// [`Event::Sliced`].
414 len: u64,
415 },
416 /// Pull one bounded window of an open file as raw bytes.
417 ///
418 /// The binary counterpart to [`Command::Slice`]. Prefer `Slice` for text:
419 /// it needs no encoding, and it handles the character-boundary problem for
420 /// you. This exists for blocks that genuinely need bytes — inspecting an
421 /// image header, say — and pays base64's cost to carry them.
422 SliceBytes {
423 /// Handle from a previous [`Command::Open`].
424 handle: Handle,
425 /// Byte offset to read from.
426 offset: u64,
427 /// Maximum bytes to return.
428 len: u64,
429 },
430 /// Extract one page of a document as text.
431 ///
432 /// Fails when the document has no text layer; check
433 /// [`MediaKind::Document::has_text_layer`] first.
434 PageText {
435 /// Handle from a previous [`Command::Open`].
436 handle: Handle,
437 /// Zero-based page number.
438 page: u32,
439 },
440 /// Render one page of a document to an image.
441 ///
442 /// Yields a *new* handle referring to the rendered image, which can then be
443 /// named in [`Command::Infer`]. That indirection is deliberate: the image
444 /// stays host-side like every other bulk value, and a rendered page is
445 /// usable exactly wherever a file-backed image is.
446 PageImage {
447 /// Handle from a previous [`Command::Open`].
448 handle: Handle,
449 /// Zero-based page number.
450 page: u32,
451 },
452 /// Embed one or more texts, returning a vector each.
453 ///
454 /// Batched by construction: the guest hands over every text it wants
455 /// embedded in one command. Embedding a corpus means tens of thousands
456 /// of chunks, and a round trip per chunk is the difference between
457 /// minutes and hours — so the batch is the primitive and a single text
458 /// is just a batch of one.
459 ///
460 /// Served by the spec's `embedding_model`, which is deliberately not the
461 /// job's chat model: they are different models, and asking a chat model
462 /// to embed either fails or returns something that is not an embedding.
463 Embed {
464 /// The texts to embed, in order. Vectors come back in the same order.
465 texts: Vec<String>,
466 },
467 /// Download a URL and hand back a handle to it.
468 ///
469 /// Answers with [`Event::Opened`], the same event [`Command::Open`]
470 /// produces, so everything downstream is unchanged: a fetched resource
471 /// can be sliced, identified, extracted from, or handed to a vision
472 /// model exactly as a local file can. One new command, no new surface.
473 ///
474 /// Requires a matching `Fetch` prefix in the spec's `capabilities`. A
475 /// corpus on the web is still a corpus, and the capability list has to
476 /// describe reaching it.
477 Fetch {
478 /// The URL to retrieve. Must begin with a granted prefix.
479 url: String,
480 },
481 /// Every character of text in a document, in one call.
482 ///
483 /// What most callers reading a PDF actually want, and the direct way to
484 /// ask for it. Before this existed the only route was
485 /// [`Command::PageText`] with `page: 0`, which reads like "the first
486 /// page" and quietly means "everything" whenever the extractor emitted
487 /// no page breaks — a gap that cost a real user two iterations and
488 /// nearly a wrong conclusion about whether cuttlefish could read their
489 /// corpus at all.
490 ///
491 /// Prefer this over walking pages unless the pages are genuinely needed
492 /// separately: the whole document is one extraction either way.
493 DocumentText {
494 /// Handle from a previous [`Command::Open`].
495 handle: Handle,
496 },
497 /// Transform an image handle, yielding a *new* image handle.
498 ///
499 /// Same indirection as [`Command::PageImage`], and for the same reason:
500 /// pixels stay host-side, and the result is usable exactly wherever a
501 /// file-backed image is — including as an [`Command::Infer`] image.
502 ///
503 /// This is the one image operation that needs a decoder, which is why it
504 /// lives here rather than in the guest. Metadata (dimensions, EXIF,
505 /// chunk structure) is header parsing and is available to any script
506 /// without a host round trip or a feature flag; *looking at the pixels*
507 /// is not.
508 ///
509 /// Hosts built without the `image-ops` feature reject this with a clear
510 /// message rather than pretending to succeed.
511 ImageOp {
512 /// Handle from a previous [`Command::Open`] or [`Command::PageImage`].
513 handle: Handle,
514 /// What to do to it.
515 op: ImageOperation,
516 },
517 /// Report progress to whoever is watching the job's event stream.
518 Emit {
519 /// Arbitrary JSON, forwarded verbatim to the job's subscribers.
520 progress: serde_json::Value,
521 },
522 /// Finish successfully with this payload.
523 Done {
524 /// The job's result, shaped by the spec's declared output.
525 result: serde_json::Value,
526 },
527 /// Give up. The job ends with this code and message, and no result.
528 Fail {
529 /// Machine-readable code; see [`error_codes`].
530 code: String,
531 /// Human-readable explanation.
532 message: String,
533 },
534}
535
536/// What the host feeds back into the guest's `step` export after carrying out a
537/// [`Command`].
538#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
539#[serde(tag = "event", rename_all = "snake_case")]
540pub enum Event {
541 /// Generation finished.
542 InferDone {
543 /// The generated text.
544 text: String,
545 /// How many tokens were produced. May be fewer than the requested
546 /// `max_tokens` if the guest ended generation early.
547 tokens_out: u32,
548 },
549 /// A file was opened.
550 Opened {
551 /// Use this in subsequent [`Command::Slice`] calls.
552 handle: Handle,
553 /// Total size of the file, in bytes.
554 len: u64,
555 /// What the host made of the contents; see [`MediaKind`].
556 ///
557 /// `#[serde(default)]` so a block built before this field existed still
558 /// deserializes, treating anything it opens as text.
559 #[serde(default)]
560 kind: MediaKind,
561 },
562 /// A window of a file was read.
563 Sliced {
564 /// The window's contents.
565 text: String,
566 /// Where the returned text actually ended.
567 ///
568 /// This is **not** always `offset + len` from the request: the host cuts
569 /// a window back to a UTF-8 character boundary, because a caller picking
570 /// window sizes has no idea where characters begin, and a naive split
571 /// would corrupt a multi-byte character at nearly every seam. A guest
572 /// walking a file must resume from this value rather than advancing by
573 /// the length it asked for.
574 next_offset: u64,
575 },
576 /// A window of a file was read as raw bytes.
577 SlicedBytes {
578 /// The window's contents, base64-encoded.
579 ///
580 /// Base64 rather than a binary side channel: the boundary is JSON, and
581 /// keeping it inspectable is worth more than the third it costs on a
582 /// path blocks are not expected to use in bulk.
583 bytes_base64: String,
584 /// Where the returned bytes ended. Unlike [`Event::Sliced`] there is no
585 /// truncation, so this is always `offset + len` clamped to the file.
586 next_offset: u64,
587 },
588 /// One vector per text handed to [`Command::Embed`], in the same order.
589 Embedded {
590 /// The vectors, each the model's full dimensionality.
591 vectors: Vec<Vec<f32>>,
592 },
593 /// Text extracted from a document — one page, or the whole thing.
594 ///
595 /// Shared by [`Command::PageText`] and [`Command::DocumentText`]: both
596 /// answer with text and nothing else, so a second variant carrying an
597 /// identical payload would add surface without adding information. The
598 /// *command* is what names the intent.
599 PageTexted {
600 /// The page's text.
601 text: String,
602 },
603 /// A document page was rendered to an image.
604 PageImaged {
605 /// A new handle referring to the rendered image; name it in
606 /// [`Command::Infer`].
607 handle: Handle,
608 /// Its size in bytes.
609 len: u64,
610 },
611 /// Progress was forwarded. Carries nothing; it exists so `Emit` has a reply
612 /// and the command loop keeps its shape.
613 Emitted,
614}
615
616/// A guest's verdict on each streamed token, returned from its `on_token`
617/// export.
618#[derive(Debug, Clone, Copy, PartialEq, Eq)]
619pub enum TokenAction {
620 /// Keep generating.
621 Continue,
622 /// Stop generating now.
623 ///
624 /// A token or two may still arrive after this, because the verdict has to
625 /// travel back to the thread doing the generating.
626 Stop,
627}
628
629impl TokenAction {
630 /// Decode the raw `i32` a guest's `on_token` export returns.
631 ///
632 /// Anything that is not an explicit `Continue` reads as `Stop`. A guest
633 /// returning a value this crate does not recognise is malfunctioning, and
634 /// the safe reading of a malfunctioning guest is "stop", never "keep
635 /// spending tokens" — the same fail-closed posture as the capability checks.
636 pub fn from_i32(v: i32) -> Self {
637 if v == 0 {
638 Self::Continue
639 } else {
640 Self::Stop
641 }
642 }
643
644 /// Encode for the wasm boundary.
645 ///
646 /// These integers are part of the ABI: renumbering them silently changes the
647 /// meaning of every already-compiled block.
648 pub fn as_i32(self) -> i32 {
649 match self {
650 Self::Continue => 0,
651 Self::Stop => 1,
652 }
653 }
654}
655
656/// Where a job is in its lifecycle.
657#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
658#[serde(rename_all = "snake_case")]
659pub enum JobStatus {
660 /// Accepted, not yet started.
661 Queued,
662 /// Executing.
663 Running,
664 /// Finished with a result.
665 Completed,
666 /// Finished with an error and no result.
667 Failed,
668 /// Stopped by request.
669 Cancelled,
670 /// Was running when the daemon last stopped; not resumed automatically.
671 Interrupted,
672}
673
674impl JobStatus {
675 /// Whether this status is final — nothing further will happen to the job.
676 ///
677 /// Clients poll until this is true. Adding a new non-terminal status is
678 /// therefore safe, while a new terminal one that is missing from this match
679 /// leaves callers waiting forever.
680 pub fn is_terminal(self) -> bool {
681 matches!(self, Self::Completed | Self::Failed | Self::Cancelled)
682 }
683}
684
685/// What a job cost.
686#[derive(Debug, Clone, Default, Serialize, Deserialize, PartialEq)]
687pub struct Usage {
688 /// Tokens consumed by prompts.
689 pub tokens_in: u32,
690 /// Tokens generated.
691 pub tokens_out: u32,
692 /// Wall-clock duration of the job.
693 pub duration_ms: u64,
694 /// Which model served the job's inference.
695 pub model: String,
696}
697
698/// The fixed, spec-independent envelope handed back to the calling agent.
699///
700/// Every job returns this shape regardless of what it did, so an agent can
701/// handle results without knowing anything about the block that produced them.
702/// Only `result` varies, and its shape is that job's business.
703#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
704pub struct Envelope {
705 /// Lifecycle state; see [`JobStatus::is_terminal`].
706 pub status: JobStatus,
707 /// Present only when the job completed.
708 ///
709 /// A failed or cancelled job never carries a partial result: a caller must
710 /// never have to guess whether a payload is trustworthy.
711 #[serde(skip_serializing_if = "Option::is_none")]
712 pub result: Option<serde_json::Value>,
713 /// Present only when the job failed or was cancelled.
714 #[serde(skip_serializing_if = "Option::is_none")]
715 pub error: Option<JobError>,
716 /// Cost accounting, populated even for failed jobs — work already spent
717 /// still counts.
718 pub usage: Usage,
719}
720
721/// Why a job did not complete.
722#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
723pub struct JobError {
724 /// Machine-readable; see [`error_codes`].
725 pub code: String,
726 /// Human-readable detail.
727 pub message: String,
728}
729
730/// The error codes the daemon emits in [`JobError::code`].
731///
732/// These are string constants rather than an enum so the set can grow without
733/// breaking clients that match on strings, and so a client built against an
734/// older version meets an unfamiliar code rather than a decode failure.
735pub mod error_codes {
736 /// The job's model could not be loaded or served.
737 pub const MODEL_LOAD_FAILED: &str = "model_load_failed";
738 /// A guest tried to reach something its spec does not grant.
739 pub const CAPABILITY_DENIED: &str = "capability_denied";
740 /// Job input did not match the spec's declared shape.
741 pub const SCHEMA_VALIDATION_FAILED: &str = "schema_validation_failed";
742 /// A Script-kind block's own logic failed at run time — a Rhai runtime
743 /// error (e.g. an out-of-bounds index, a missing function) surfacing
744 /// from inside the script itself. `catalog add` rejects a script whose
745 /// body doesn't *parse*, so this is specifically a failure that only
746 /// shows up once the script actually runs; it is not a schema mismatch,
747 /// so it does not reuse [`SCHEMA_VALIDATION_FAILED`].
748 pub const SCRIPT_ERROR: &str = "script_error";
749 /// The guest trapped — a panic, a bad export signature, or malformed wasm.
750 pub const WASM_TRAP: &str = "wasm_trap";
751 /// The job exceeded its time budget.
752 pub const TIMEOUT: &str = "timeout";
753 /// The job was cancelled by request.
754 pub const CANCELLED: &str = "cancelled";
755 /// A command needed a capability this build does not have — asking for a
756 /// page image without document rendering compiled in, say.
757 pub const UNSUPPORTED: &str = "unsupported";
758 /// A path the job may read names nothing.
759 ///
760 /// Distinct from [`CAPABILITY_DENIED`] on purpose: one is a wrong path
761 /// and the other a wrong grant, and they are fixed in different files.
762 /// Only used where the job could have discovered the absence itself —
763 /// outside a grant, absence and refusal stay indistinguishable, because
764 /// telling them apart is an existence oracle.
765 pub const NOT_FOUND: &str = "not_found";
766}
767
768impl Default for MediaKind {
769 /// Text, because that is what every command predating [`MediaKind`]
770 /// assumed, and because it keeps an older block's behaviour unchanged.
771 fn default() -> Self {
772 Self::Text
773 }
774}
775
776#[cfg(test)]
777mod tests {
778 use super::*;
779
780 #[test]
781 fn a_signature_round_trips_through_its_compact_string() {
782 let sig = Signature {
783 input: Ty::Record([("path".to_string(), Ty::Text)].into_iter().collect()),
784 output: Ty::List(Box::new(Ty::Text)),
785 };
786 let s = sig.to_string();
787 assert_eq!(s, "{path: text} -> [text]");
788 assert_eq!(s.parse::<Signature>().unwrap(), sig);
789 }
790
791 #[test]
792 fn a_signature_without_an_arrow_is_rejected() {
793 assert!("just-a-type".parse::<Signature>().is_err());
794 }
795
796 #[test]
797 fn a_signature_with_an_unparseable_side_is_rejected() {
798 assert!("text -> not a type".parse::<Signature>().is_err());
799 }
800
801 #[test]
802 fn text_matches_a_json_string_only() {
803 assert!(Ty::Text.matches_value(&serde_json::json!("hello")));
804 assert!(!Ty::Text.matches_value(&serde_json::json!(42)));
805 }
806
807 #[test]
808 fn json_matches_anything() {
809 assert!(Ty::Json.matches_value(&serde_json::json!(null)));
810 assert!(Ty::Json.matches_value(&serde_json::json!([1, "two", {}])));
811 }
812
813 #[test]
814 fn a_record_missing_a_declared_field_does_not_match() {
815 let ty = Ty::Record([("summary".to_string(), Ty::Text)].into_iter().collect());
816 assert!(!ty.matches_value(&serde_json::json!({"text": "hello world"})));
817 }
818
819 #[test]
820 fn a_record_with_the_declared_field_present_and_well_typed_matches() {
821 let ty = Ty::Record([("summary".to_string(), Ty::Text)].into_iter().collect());
822 assert!(ty.matches_value(&serde_json::json!({"summary": "hi", "extra": 1})));
823 }
824
825 #[test]
826 fn a_record_with_a_wrong_typed_declared_field_does_not_match() {
827 let ty = Ty::Record([("summary".to_string(), Ty::Text)].into_iter().collect());
828 assert!(!ty.matches_value(&serde_json::json!({"summary": 42})));
829 }
830
831 #[test]
832 fn a_list_matches_only_when_every_element_matches_the_inner_type() {
833 let ty = Ty::List(Box::new(Ty::Text));
834 assert!(ty.matches_value(&serde_json::json!(["a", "b"])));
835 assert!(!ty.matches_value(&serde_json::json!(["a", 2])));
836 assert!(!ty.matches_value(&serde_json::json!("not a list")));
837 }
838
839 #[test]
840 fn bytes_image_and_document_accept_anything() {
841 for ty in [Ty::Bytes, Ty::Image, Ty::Document] {
842 assert!(ty.matches_value(&serde_json::json!(123)));
843 }
844 }
845}