scriptoriumadmin/fabro
1
0
Fork 0
mirror of https://github.com/fabro-sh/fabro.git synced 2026-10-01 02:04:24 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Merge pull request #883 from fabro-sh/bump-lithos-llm-codecs

Browse source 
Bump lithos-llm to 43a42ac and migrate catalogs to the codecs schema
This commit is contained in:
Bryan Helmkamp 2026-09-18 22:23:29 -04:00 • committed by GitHub
commit f3df72e089
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
29 changed files with 112 additions and 108 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

31
Cargo.lock generated
View file

@ -2063,7 +2063,7 @@ dependencies = [
"libc",
"option-ext",
"redox_users",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -2190,7 +2190,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
dependencies = [
"libc",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -4358,7 +4358,7 @@ dependencies = [
"js-sys",
"log",
"wasm-bindgen",
"windows-core 0.61.2",
"windows-core 0.62.2",
]
[[package]]
@ -4888,7 +4888,7 @@ checksum = "6373607a59f0be73a39b6fe456b8192fcc3585f602af20751600e974dd455e77"
[[package]]
name = "lithos-llm"
version = "0.1.0"
source = "git+https://github.com/lithoscomputer/lithos-llm?rev=55add4596b861a0623d00c3a54aa5c147c8d504b#55add4596b861a0623d00c3a54aa5c147c8d504b"
source = "git+https://github.com/lithoscomputer/lithos-llm?rev=43a42ac28e9d9bcf40a91abc02be4f12ca274ebb#43a42ac28e9d9bcf40a91abc02be4f12ca274ebb"
dependencies = [
"async-trait",
"aws-config",
@ -4900,6 +4900,7 @@ dependencies = [
"crc32fast",
"futures-core",
"futures-util",
"indexmap 2.13.0",
"mime_guess",
"reqwest 0.13.4",
"serde",
@ -5304,7 +5305,7 @@ version = "0.50.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7957b9740744892f114936ab4a57b3f487491bbeafaf8083688b16841a4240e5"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -5869,7 +5870,7 @@ dependencies = [
[[package]]
name = "pebble-agent"
version = "0.1.0"
source = "git+https://github.com/lithoscomputer/pebble?rev=a39f43e26effdf99635eaf343f095c17157c9c93#a39f43e26effdf99635eaf343f095c17157c9c93"
source = "git+https://github.com/lithoscomputer/pebble?rev=67c9f486dd28f15c04e8d590a91e6f5563f7605d#67c9f486dd28f15c04e8d590a91e6f5563f7605d"
dependencies = [
"async-trait",
"futures-util",
@ -5886,7 +5887,7 @@ dependencies = [
[[package]]
name = "pebble-cli-core"
version = "0.1.0"
source = "git+https://github.com/lithoscomputer/pebble?rev=a39f43e26effdf99635eaf343f095c17157c9c93#a39f43e26effdf99635eaf343f095c17157c9c93"
source = "git+https://github.com/lithoscomputer/pebble?rev=67c9f486dd28f15c04e8d590a91e6f5563f7605d#67c9f486dd28f15c04e8d590a91e6f5563f7605d"
dependencies = [
"anyhow",
"async-trait",
@ -5915,7 +5916,7 @@ dependencies = [
[[package]]
name = "pebble-coding-agent"
version = "0.1.0"
source = "git+https://github.com/lithoscomputer/pebble?rev=a39f43e26effdf99635eaf343f095c17157c9c93#a39f43e26effdf99635eaf343f095c17157c9c93"
source = "git+https://github.com/lithoscomputer/pebble?rev=67c9f486dd28f15c04e8d590a91e6f5563f7605d#67c9f486dd28f15c04e8d590a91e6f5563f7605d"
dependencies = [
"async-trait",
"futures-util",
@ -6284,7 +6285,7 @@ dependencies = [
"once_cell",
"socket2",
"tracing",
"windows-sys 0.60.2",
"windows-sys 0.59.0",
]
[[package]]
@ -6761,7 +6762,7 @@ dependencies = [
"errno 0.3.14",
"libc",
"linux-raw-sys",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -6820,7 +6821,7 @@ dependencies = [
"security-framework",
"security-framework-sys",
"webpki-root-certs",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -7467,7 +7468,7 @@ version = "1.4.8"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c4db69cba1110affc0e9f7bcd48bbf87b3f4fc7c61fc9155afd4c469eb3d6c1b"
dependencies = [
"errno 0.2.8",
"errno 0.3.14",
"libc",
]
@ -8032,7 +8033,7 @@ dependencies = [
"getrandom 0.4.1",
"once_cell",
"rustix",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -8067,7 +8068,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "230a1b821ccbd75b185820a1f1ff7b14d21da1e442e22c0863ea5f08771a8874"
dependencies = [
"rustix",
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]
@ -9134,7 +9135,7 @@ version = "0.1.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
dependencies = [
"windows-sys 0.61.2",
"windows-sys 0.59.0",
]
[[package]]

8
Cargo.toml
View file

@ -93,7 +93,7 @@ insta = "1"
fabro-test = { path = "lib/foundation/fabro-test" }
# Provider-neutral LLM catalog and client. Pinned to a revision until 0.x is
# published to crates.io.
lithos-llm = { git = "https://github.com/lithoscomputer/lithos-llm", rev = "55add4596b861a0623d00c3a54aa5c147c8d504b", default-features = false }
lithos-llm = { git = "https://github.com/lithoscomputer/lithos-llm", rev = "43a42ac28e9d9bcf40a91abc02be4f12ca274ebb", default-features = false }
# Deterministic OpenAI twin used by twin-mode E2E tests; the same revision
# lithos-llm verifies its codecs against.
twin-openai = { git = "https://github.com/lithoscomputer/twins", rev = "ca45f0e50a6716d716aa2f638ca3cf767e88f613" }
@ -124,9 +124,9 @@ sandbox-driver-testing = { git = "https://github.com/lithoscomputer/sandbox-driv
# sandbox, so the pebble and sandbox-driver pins move independently. Pebble
# pins the same lithos-llm rev as fabro, and its lockfile policy is that
# every shared crate resolves to the version lithos-llm locks.
pebble-agent = { git = "https://github.com/lithoscomputer/pebble", rev = "a39f43e26effdf99635eaf343f095c17157c9c93" }
pebble-coding-agent = { git = "https://github.com/lithoscomputer/pebble", rev = "a39f43e26effdf99635eaf343f095c17157c9c93", features = ["mcp", "search-providers"] }
pebble-cli-core = { git = "https://github.com/lithoscomputer/pebble", rev = "a39f43e26effdf99635eaf343f095c17157c9c93" }
pebble-agent = { git = "https://github.com/lithoscomputer/pebble", rev = "67c9f486dd28f15c04e8d590a91e6f5563f7605d" }
pebble-coding-agent = { git = "https://github.com/lithoscomputer/pebble", rev = "67c9f486dd28f15c04e8d590a91e6f5563f7605d", features = ["mcp", "search-providers"] }
pebble-cli-core = { git = "https://github.com/lithoscomputer/pebble", rev = "67c9f486dd28f15c04e8d590a91e6f5563f7605d" }
sentry = { version = "0.35", default-features = false, features = ["backtrace", "contexts", "ureq", "rustls"] }
fork = "0.2"
exec = "0.3"

4
docs/public/api-reference/fabro-api.yaml
View file

@ -8417,8 +8417,8 @@ components:
example: "Anthropic"
adapter:
type: string
description: "lithos adapter id the provider speaks, such as `anthropic`, `openai`, `gemini`, or `openai-compatible`."
example: "anthropic"
description: "lithos adapter id the provider uses: `http` or `bedrock`, or a custom adapter's id."
example: "http"
base_url:
type: string
description: Effective API base URL, including any operator override.

4
docs/public/core-concepts/models.mdx
View file

@ -80,13 +80,11 @@ Claude Fable 5 is available as an explicit model but is not the default Anthropi
Fabro's catalog is the [lithos-llm](https://docs.rs/lithos-llm) built-in catalog. The `[llm]` table in settings is a second layer over it: a lithos catalog overlay that adds providers and models or changes existing entries. Later layers win. Tables merge key by key and every other value replaces. Models are nested under their provider, so two providers can expose the same model id without overwriting each other.
Provider and model facts use lithos field names: `adapter`, `codec`, `base_url`, `auth`, `enabled`, `limits`, `capabilities`, `pricing`, `small_default`, `probe`, `family`, and the cutoffs. The coding harness a model expects lives under `metadata.agent`, a namespace lithos ships and other agents such as Pebble read too. See [Settings Configuration](/reference/user-configuration#llm) for every key.
Provider and model facts use lithos field names: `adapter`, `codecs`, `base_url`, `auth`, `enabled`, `limits`, `capabilities`, `pricing`, `small_default`, `probe`, `family`, and the cutoffs. The coding harness a model expects lives under `metadata.agent`, a namespace lithos ships and other agents such as Pebble read too. See [Settings Configuration](/reference/user-configuration#llm) for every key.
```toml title="settings.toml"
[llm.providers.proxy]
display_name = "Acme Gateway"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://llm-gateway.example.com/v1"
auth = { type = "bearer" }
aliases = ["gateway"]

2
docs/public/integrations/fireworks.mdx
View file

@ -3,7 +3,7 @@ title: "Fireworks AI"
description: "Run open-weights models on Fireworks AI's serverless inference platform"
---
[Fireworks AI](https://fireworks.ai/) serves open-weights models (Kimi, DeepSeek, GLM, Qwen, GPT-OSS, and more) behind an OpenAI-compatible API. Fabro ships a disabled `fireworks` provider entry with a curated model catalog, so you can opt in from `settings.toml` without changing Fabro code.
[Fireworks AI](https://fireworks.ai/) serves open-weights models (Kimi, DeepSeek, GLM, Qwen, GPT-OSS, and more) behind an OpenAI-compatible API. Fabro ships an enabled `fireworks` provider entry with a curated model catalog; it needs only an API key, and `settings.toml` can adjust it without changing Fabro code.
## Prerequisites

2
docs/public/integrations/openrouter.mdx
View file

@ -3,7 +3,7 @@ title: "OpenRouter"
description: "Route Fabro models through OpenRouter's multi-provider gateway"
---
[OpenRouter](https://openrouter.ai/) is an aggregator that fronts hundreds of models behind one OpenAI-compatible API. Fabro ships a disabled `openrouter` provider entry with a curated model catalog, so you can opt in from `settings.toml` without changing Fabro code.
[OpenRouter](https://openrouter.ai/) is an aggregator that fronts hundreds of models behind one OpenAI-compatible API. Fabro ships an enabled `openrouter` provider entry with a curated model catalog; it needs only an API key, and `settings.toml` can adjust it without changing Fabro code.
## Prerequisites

2
docs/public/reference/sdk.mdx
View file

@ -449,7 +449,7 @@ let result = client.complete_with_context(request, context).await;
### Provider adapters
Providers are lithos adapters selected by the catalog `adapter` id: `anthropic`, `openai`, `gemini`, `openai-compatible`, and `bedrock`. A new OpenAI-compatible endpoint needs a catalog entry, not code.
A provider names one lithos adapter in `adapter` (`http`, the default, or `bedrock`) and lists the wire codecs its host speaks in `codecs` (`openai-chat`, the default, `openai-responses`, `anthropic-messages`, `gemini-generate`, or `bedrock-converse`). A new OpenAI-compatible endpoint needs a catalog entry, not code.
To add a custom transport, implement the lithos `ProviderAdapter` trait and register it with `ClientOptions::with_adapter`. `fabro_llm::gateway::GatewayAdapter` is Fabro's own example: it posts each request to a Fabro server's completions endpoint, which returns lithos `Response` JSON and streams lithos `StreamEvent` JSON verbatim.

12
docs/public/reference/user-configuration.mdx
View file

@ -89,8 +89,6 @@ level = "info"
[llm.providers.proxy]
display_name = "Acme Gateway"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://llm-gateway.example.com/v1"
auth = { type = "bearer" }
aliases = ["gateway"]
@ -154,8 +152,6 @@ Several built-in providers ship with `enabled = false`. Turn one on by setting
```toml title="settings.toml"
[llm.providers.proxy]
display_name = "Acme Gateway"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://llm-gateway.example.com/v1"
auth = { type = "bearer" }
priority = 50
@ -195,11 +191,11 @@ Define or override an LLM provider. The keys are the lithos provider record.
| Key | Type / values | Default | Description |
|---|---|---|---|
| `display_name` | string | required for new providers | Human-readable provider name. |
| `adapter` | string | required for new providers | lithos adapter id: `anthropic`, `openai`, `gemini`, `openai-compatible`, or `bedrock`. |
| `codec` | string | required for new providers | Wire codec: `anthropic-messages`, `openai-responses`, `openai-chat`, `gemini-generate`, or `bedrock-converse`. |
| `base_url` | string | required for new providers | Provider API base URL. The `openai-compatible` adapter appends `/v1/chat/completions` unless the URL already ends in a version segment. |
| `adapter` | string | `"http"` | lithos adapter id: `http` or `bedrock`. Any other id names a custom adapter the application registered. |
| `codecs` | array<string> | `["openai-chat"]` | Wire codecs the host speaks, in preference order: `openai-chat`, `openai-responses`, `anthropic-messages`, `gemini-generate`, or `bedrock-converse`. A Chat Completions host needs no line. |
| `base_url` | string | required for new providers | Provider API base URL. The `openai-chat` codec appends `/v1/chat/completions` unless the URL already ends in a version segment. |
| `auth` | table | required for new providers | Auth scheme: `{ type = "bearer" }`, `{ type = "header", name = "x-api-key" }`, `{ type = "headers" }`, `{ type = "none" }`, or `{ type = "aws" }`. |
| `enabled` | boolean | `true` | Set `false` to hide a provider from Fabro. `bedrock`, `bedrock-openai`, `fireworks`, `litellm`, `modal`, `ollama`, and `openrouter` ship disabled. |
| `enabled` | boolean | `true` | Set `false` to hide a provider from Fabro. `bedrock`, `bedrock-openai`, `litellm`, `modal`, and `ollama` ship disabled. |
| `priority` | integer | `0` | Higher-priority ready providers win unqualified model and default selection. |
| `aliases` | array<string> | `[]` | Additional provider names accepted by model routing and fallback config. |
| `default_model` | string | None | The provider's default model id. |

4
lib/apps/fabro-cli/tests/it/cmd/exec.rs
View file

@ -333,7 +333,7 @@ fn exec_accepts_configured_custom_provider_from_settings() {
let context = test_context!();
context.write_home(
".fabro/settings.toml",
"_version = 1\n\n[llm.providers.acme-aws]\ndisplay_name = \"Acme AWS\"\nadapter = \"openai-compatible\"\ncodec = \"openai-chat\"\nbase_url = \"https://bedrock.example.invalid/v1\"\nauth = { type = \"bearer\" }\nallow_passthrough = true\n\n[llm.providers.acme-aws.metadata.agent]\nprofile = \"openai\"\n\n[cli.exec.model]\nprovider = \"acme-aws\"\nname = \"acme-claude-sonnet-4-6\"\n",
"_version = 1\n\n[llm.providers.acme-aws]\ndisplay_name = \"Acme AWS\"\nbase_url = \"https://bedrock.example.invalid/v1\"\nauth = { type = \"bearer\" }\nallow_passthrough = true\n\n[llm.providers.acme-aws.metadata.agent]\nprofile = \"openai\"\n\n[cli.exec.model]\nprovider = \"acme-aws\"\nname = \"acme-claude-sonnet-4-6\"\n",
);
let mut cmd = context.exec_cmd();
@ -398,7 +398,7 @@ fn exec_server_target_accepts_configured_custom_provider_from_settings() {
let context = test_context!();
context.write_home(
".fabro/settings.toml",
"_version = 1\n\n[llm.providers.acme-aws]\ndisplay_name = \"Acme AWS\"\nadapter = \"openai-compatible\"\ncodec = \"openai-chat\"\nbase_url = \"https://bedrock.example.invalid/v1\"\nauth = { type = \"bearer\" }\nallow_passthrough = true\n\n[llm.providers.acme-aws.metadata.agent]\nprofile = \"openai\"\n\n[cli.exec.model]\nprovider = \"acme-aws\"\nname = \"acme-claude-sonnet-4-6\"\n",
"_version = 1\n\n[llm.providers.acme-aws]\ndisplay_name = \"Acme AWS\"\nbase_url = \"https://bedrock.example.invalid/v1\"\nauth = { type = \"bearer\" }\nallow_passthrough = true\n\n[llm.providers.acme-aws.metadata.agent]\nprofile = \"openai\"\n\n[cli.exec.model]\nprovider = \"acme-aws\"\nname = \"acme-claude-sonnet-4-6\"\n",
);
let server = MockServer::start();
server.mock(|when, then| {

2
lib/apps/fabro-server/src/run_manifest.rs
View file

@ -3039,8 +3039,6 @@ digraph Demo {
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.acme.test/v1"
auth = { type = "bearer" }
default_model = "acme-large"

4
lib/apps/fabro-server/src/server/handler/sessions.rs
View file

@ -1530,8 +1530,8 @@ mod tests {
}
/// OpenAI and OpenRouter both offer `gpt-5.6-sol` under the `gpt-56-sol`
/// alias; OpenRouter ships disabled, so enable it the way an operator
/// would.
/// alias; both ship enabled, and the overlay makes it the default on
/// each.
fn portable_session_catalog() -> Catalog {
fabro_llm::test_support::test_catalog_with_overlay(
r#"

18
lib/apps/fabro-server/src/server/tests.rs
View file

@ -335,8 +335,6 @@ fn acme_overlay(base_url: &str) -> String {
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = {base_url}
auth = {{ type = "bearer" }}
priority = 120
@ -8417,8 +8415,6 @@ async fn model_api_keeps_duplicate_ids_provider_scoped_and_selects_ready_priorit
r#"
[providers.direct]
display_name = "Direct"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = {direct}
auth = {{ type = "bearer" }}
priority = 120
@ -8436,8 +8432,6 @@ capabilities = {{ text = true }}
[providers.aggregator]
display_name = "Aggregator"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = {aggregator}
auth = {{ type = "bearer" }}
priority = 110
@ -8600,8 +8594,6 @@ async fn test_model_forwards_and_validates_reasoning_effort() {
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = {base_url}
auth = {{ type = "bearer" }}
priority = 120
@ -9152,8 +9144,8 @@ async fn test_providers_registration_issue_returns_error_without_probe() {
// An adapter lithos does not ship cannot be built, so the provider is
// configured (it has a vault key) yet unavailable.
let overlay = acme_overlay("https://api.acme.test/v1").replace(
"adapter = \"openai-compatible\"",
"adapter = \"not-an-adapter\"",
"display_name = \"Acme\"",
"display_name = \"Acme\"\nadapter = \"not-an-adapter\"",
);
let state = TestAppStateBuilder::new()
.runtime_settings(default_test_server_settings(), RunLayer::default())
@ -9226,8 +9218,7 @@ async fn test_providers_mixed_results_preserve_catalog_order_and_counts() {
r#"
[providers.zeta]
display_name = "Zeta"
adapter = "openai"
codec = "openai-responses"
codecs = ["openai-responses"]
base_url = {base_url}
auth = {{ type = "bearer" }}
priority = 50
@ -9242,8 +9233,7 @@ probe = true
[providers.alpha]
display_name = "Alpha"
adapter = "openai"
codec = "openai-responses"
codecs = ["openai-responses"]
base_url = {base_url}
auth = {{ type = "bearer" }}
priority = 40

35
lib/apps/fabro-server/tests/it/pagination.rs
View file

@ -72,16 +72,31 @@ async fn paginated_endpoints_return_correct_shape() {
let app = fabro_server::test_support::build_test_router(state);
for ep in ENDPOINTS {
// Large limit: paginated shape, has_more = false (all fixture items fit).
// Using an explicit large limit instead of the server default so the test
// stays robust when datasets (e.g. the built-in model catalog) grow.
let json = get_json(app.clone(), &format!("{}?page[limit]=100", ep.path)).await;
assert_paginated_shape(&json, ep.name);
assert_eq!(
json["meta"]["has_more"], false,
"{}: large limit should have has_more=false",
ep.name
);
// Walk the collection at the largest page the API allows: every page
// has the paginated shape, and the last one reports has_more = false.
// The built-in model catalog is larger than one page, so the walk
// follows `page[offset]` rather than assuming one page fits.
let mut offset = 0;
loop {
let json = get_json(
app.clone(),
&format!("{}?page[limit]=100&page[offset]={offset}", ep.path),
)
.await;
assert_paginated_shape(&json, &format!("{} offset={offset}", ep.name));
let page_len = json["data"].as_array().unwrap().len();
assert!(page_len <= 100, "{}: page exceeded the limit", ep.name);
if json["meta"]["has_more"] == false {
break;
}
assert!(
page_len == 100,
"{}: has_more=true on a page shorter than the limit",
ep.name
);
offset += page_len;
assert!(offset < 10_000, "{}: has_more never turned false", ep.name);
}
// limit=1: at most 1 item, has_more = true (all fixtures have >1 item)
let json = get_json(app.clone(), &format!("{}?page[limit]=1", ep.path)).await;

2
lib/components/fabro-llm/Cargo.toml
View file

@ -27,7 +27,7 @@ fabro-redact.workspace = true
fabro-static.workspace = true
fabro-types = { path = "../../foundation/fabro-types" }
futures.workspace = true
lithos-llm = { workspace = true, features = ["builtin-catalog", "openai", "anthropic", "gemini", "openai-compatible", "bedrock", "bedrock-aws", "local-files"] }
lithos-llm = { workspace = true, features = ["builtin-catalog", "bedrock", "bedrock-aws", "local-files"] }
serde.workspace = true
serde_json.workspace = true
strum.workspace = true

43
lib/components/fabro-llm/src/catalog.rs
View file

@ -13,7 +13,8 @@ use fabro_static::EnvVars;
use fabro_types::AgentProfileKind;
pub use lithos_llm::catalog::Offering;
use lithos_llm::catalog::{
Catalog, CatalogBuilder, CatalogError, CatalogModel, CatalogProvider, Metadata,
Catalog, CatalogBuilder, CatalogError, CatalogModel, CatalogProvider, CodecId, Metadata,
adapter_ids, codec_ids,
};
use serde::Deserialize;
@ -101,11 +102,16 @@ fn implied_agent_profiles(catalog: &Catalog) -> String {
}
/// The profile a provider's wire protocol implies, for a provider whose
/// catalog entry does not name one.
/// catalog entry does not name one. The protocol is the provider's first
/// codec, the one the client sends a generation call on; the `bedrock`
/// adapter speaks Converse to Anthropic-shaped models.
fn adapter_agent_profile(provider: &CatalogProvider) -> AgentProfileKind {
match provider.adapter().as_str() {
"anthropic" | "bedrock" => AgentProfileKind::Anthropic,
"gemini" => AgentProfileKind::Gemini,
if provider.adapter().as_str() == adapter_ids::BEDROCK {
return AgentProfileKind::Anthropic;
}
match provider.codecs().first().map(CodecId::as_str) {
Some(codec_ids::ANTHROPIC_MESSAGES) => AgentProfileKind::Anthropic,
Some(codec_ids::GEMINI_GENERATE) => AgentProfileKind::Gemini,
_ => AgentProfileKind::OpenAi,
}
}
@ -241,7 +247,7 @@ enabled = false
Some(AgentProfileKind::OpenAi)
);
assert_eq!(
agent_profile(&catalog, "openrouter", None),
agent_profile(&catalog, "bedrock-openai", None),
None,
"disabled providers have no profile to offer"
);
@ -282,8 +288,6 @@ enabled = false
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.acme.test/v1"
auth = { type = "bearer" }
default_model = "acme-llama"
@ -308,4 +312,27 @@ capabilities = { text = true, tools = true }
Some(AgentProfileKind::OpenAi)
);
}
/// The implied profile follows the provider's first codec, so a host that
/// speaks Anthropic Messages gets the Anthropic harness.
#[test]
fn the_implied_profile_follows_the_first_codec() {
let overlay = LlmLayer(
toml::from_str(
r#"
[providers.acme]
display_name = "Acme"
codecs = ["anthropic-messages"]
base_url = "https://api.acme.test"
auth = { type = "header", name = "x-api-key" }
"#,
)
.unwrap(),
);
let catalog = build_catalog(&overlay, &|_| None).unwrap();
assert_eq!(
agent_profile(&catalog, "acme", None),
Some(AgentProfileKind::Anthropic)
);
}
}

8
lib/components/fabro-llm/src/selection.rs
View file

@ -407,12 +407,12 @@ mod tests {
fn disabled_providers_are_not_selectable() {
let catalog = test_catalog();
assert!(matches!(
select(&catalog, "gpt-5.4", None, &eligible(&["openrouter"])),
select(&catalog, "gpt-5.4", None, &eligible(&["bedrock-openai"])),
Err(ModelSelectionError::NoEligibleOffering { .. })
));
let enabled = test_catalog_with_overlay("[providers.openrouter]\nenabled = true\n");
let entry = select(&enabled, "gpt-5.4", None, &eligible(&["openrouter"])).unwrap();
assert_eq!(entry.provider.id(), &ProviderId::new("openrouter"));
let enabled = test_catalog_with_overlay("[providers.bedrock-openai]\nenabled = true\n");
let entry = select(&enabled, "gpt-5.4", None, &eligible(&["bedrock-openai"])).unwrap();
assert_eq!(entry.provider.id(), &ProviderId::new("bedrock-openai"));
}
#[test]

2
lib/components/fabro-validate/src/lib.rs
View file

@ -215,8 +215,6 @@ mod tests {
r#"
[providers.acme-venice]
display_name = "Acme Venice"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.venice.ai/api/v1"
auth = { type = "bearer" }
default_model = "venice-large"

5
lib/components/fabro-workflow/src/handler/llm/fallback.rs
View file

@ -286,8 +286,9 @@ mod tests {
use super::*;
/// Modal and OpenRouter ship disabled; enable them the way an operator
/// would so their models become fallback targets.
/// Modal ships disabled; enable it the way an operator would so its
/// models become fallback targets. OpenRouter ships enabled; the line
/// is kept so the overlay reads the same for both.
fn enabled_fallback_catalog() -> Catalog {
test_catalog_with_overlay(
"[providers.modal]\nenabled = true\n\n[providers.openrouter]\nenabled = true\n",

4
lib/components/fabro-workflow/src/handler/prompt.rs
View file

@ -680,8 +680,6 @@ mod tests {
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.acme.test/v1"
auth = { type = "bearer" }
default_model = "acme-claude"
@ -756,8 +754,6 @@ mod tests {
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.acme.test/v1"
auth = { type = "bearer" }
default_model = "acme-claude"

2
lib/components/fabro-workflow/src/operations/start.rs
View file

@ -1481,8 +1481,6 @@ mod tests {
r#"
[providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.acme.test/v1"
auth = { type = "bearer" }
default_model = "acme-claude"

2
lib/components/fabro-workflow/src/pipeline/pull_request.rs
View file

@ -779,8 +779,6 @@ mod tests {
r#"
[providers.mock]
display_name = "Mock"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "http://mock.invalid/v1"
auth = { type = "bearer" }
allow_passthrough = true

2
lib/components/fabro-workflow/src/transforms/model_resolution.rs
View file

@ -167,8 +167,6 @@ mod tests {
r#"
[providers.acme-venice]
display_name = "Acme Venice"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.venice.ai/api/v1"
auth = { type = "bearer" }
priority = 200

2
lib/components/fabro-workflow/tests/it/integration.rs
View file

@ -2682,8 +2682,6 @@ async fn shared_thread_compaction_before_routing_audit_succeeds() {
r#"
[providers.compact]
display_name = "Compact"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = {base_url}
auth = {{ type = "bearer" }}
default_model = "compact-model"

2
lib/components/fabro-workflow/tests/it/pebble_agent.rs
View file

@ -133,8 +133,6 @@ fn provider_toml(name: &str, model: &str, base_url: &str, profile: &str) -> Stri
r#"
[providers.{name}]
display_name = "{name}"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = {base_url}
auth = {{ type = "bearer" }}
default_model = "{model}"

2
lib/foundation/fabro-auth/src/vault_source.rs
View file

@ -383,8 +383,6 @@ mod tests {
r#"
[providers.gateway]
display_name = "Gateway"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://gateway.test/v1"
auth = { type = "bearer" }
default_headers = { "x-portkey-api-key" = "{{ secrets.PORTKEY_API_KEY }}", "x-portkey-config" = "@prod" }

2
lib/foundation/fabro-config/src/builders.rs
View file

@ -690,8 +690,6 @@ methods = ["dev-token"]
[llm.providers.acme]
display_name = "Acme"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://api.acme.test/v1"
auth = { type = "bearer" }
enabled = true

10
lib/foundation/fabro-dev/src/commands/docs_options_reference.rs
View file

@ -229,8 +229,6 @@ Several built-in providers ship with `enabled = false`. Turn one on by setting
```toml title="settings.toml"
[llm.providers.proxy]
display_name = "Acme Gateway"
adapter = "openai-compatible"
codec = "openai-chat"
base_url = "https://llm-gateway.example.com/v1"
auth = { type = "bearer" }
priority = 50
@ -270,11 +268,11 @@ Define or override an LLM provider. The keys are the lithos provider record.
| Key | Type / values | Default | Description |
|---|---|---|---|
| `display_name` | string | required for new providers | Human-readable provider name. |
| `adapter` | string | required for new providers | lithos adapter id: `anthropic`, `openai`, `gemini`, `openai-compatible`, or `bedrock`. |
| `codec` | string | required for new providers | Wire codec: `anthropic-messages`, `openai-responses`, `openai-chat`, `gemini-generate`, or `bedrock-converse`. |
| `base_url` | string | required for new providers | Provider API base URL. The `openai-compatible` adapter appends `/v1/chat/completions` unless the URL already ends in a version segment. |
| `adapter` | string | `"http"` | lithos adapter id: `http` or `bedrock`. Any other id names a custom adapter the application registered. |
| `codecs` | array<string> | `["openai-chat"]` | Wire codecs the host speaks, in preference order: `openai-chat`, `openai-responses`, `anthropic-messages`, `gemini-generate`, or `bedrock-converse`. A Chat Completions host needs no line. |
| `base_url` | string | required for new providers | Provider API base URL. The `openai-chat` codec appends `/v1/chat/completions` unless the URL already ends in a version segment. |
| `auth` | table | required for new providers | Auth scheme: `{ type = "bearer" }`, `{ type = "header", name = "x-api-key" }`, `{ type = "headers" }`, `{ type = "none" }`, or `{ type = "aws" }`. |
| `enabled` | boolean | `true` | Set `false` to hide a provider from Fabro. `bedrock`, `bedrock-openai`, `fireworks`, `litellm`, `modal`, `ollama`, and `openrouter` ship disabled. |
| `enabled` | boolean | `true` | Set `false` to hide a provider from Fabro. `bedrock`, `bedrock-openai`, `litellm`, `modal`, and `ollama` ship disabled. |
| `priority` | integer | `0` | Higher-priority ready providers win unqualified model and default selection. |
| `aliases` | array<string> | `[]` | Additional provider names accepted by model routing and fallback config. |
| `default_model` | string | None | The provider's default model id. |

2
lib/foundation/fabro-types/src/catalog_api.rs
View file

@ -75,7 +75,7 @@ pub struct Model {
pub struct Provider {
pub id: ProviderId,
pub display_name: String,
/// lithos adapter id, such as `openai` or `openai-compatible`.
/// lithos adapter id: `http` or `bedrock`, or a custom adapter's id.
pub adapter: String,
pub base_url: String,
#[serde(default, skip_serializing_if = "Option::is_none")]

2
lib/packages/fabro-api-client/src/models/provider.ts generated
View file

@ -27,7 +27,7 @@ export interface Provider {
*/
'display_name': string;
/**
* lithos adapter id the provider speaks, such as `anthropic`, `openai`, `gemini`, or `openai-compatible`.
* lithos adapter id the provider uses: `http` or `bedrock`, or a custom adapter's id.
*/
'adapter': string;
/**