Adopt daemon-style inference config for v0.1.1

Author: danias

CHANGELOG.md

@@ -0,0 +1,33 @@
+## [0.1.1] - 2026-04-14
+
+### Added
+
+- Support for reading text-generation profiles from the Bitloops daemon config schema, including shared runtimes under `[inference.runtimes.<name>]`.
+- Validation for legacy profile fields, unsupported profile keys, missing runtime references, invalid numeric values, and Ollama chat URLs that do not target `/api/chat`.
+
+### Changed
+
+- `bitloops-inference` now reads `task`, `driver`, and `runtime` from daemon-style profile definitions instead of the older per-profile `kind`, `provider_name`, and `timeout_secs` fields.
+- Non-`text_generation` inference profiles are ignored during config loading and profile discovery, so CLI commands only expose runnable text-generation profiles.
+- Documentation and test fixtures now use the daemon config layout and string-based temperature examples with environment interpolation.
+
+### Fixed
+
+- Request timeouts are now resolved from the referenced runtime’s `request_timeout_secs` value instead of per-profile timeout settings.
+- Config validation now reports when a config file does not define any text-generation profiles.
+
+## [0.1.0] - 2026-04-13
+
+### Added
+
+- Initial `bitloops-inference` Rust workspace with a shared protocol crate and a stdio runtime for out-of-process Bitloops inference.
+- Protocol v1 request and response types for `describe`, `infer`, and `shutdown`, using line-delimited JSON over `stdin` and `stdout`.
+- `run`, `validate-config`, and `describe-profile` CLI commands for running the runtime and inspecting configured inference profiles.
+- OpenAI Chat Completions and Ollama Chat providers with normalised text and `json_object` responses, usage reporting, finish reasons, and provider-specific HTTP error handling.
+- TOML-based profile configuration with environment-variable interpolation and default inference settings for temperature and output token limits.
+- Mocked provider integration tests, child-process protocol-loop tests, hosted-runner CI, and release automation for macOS, Linux, and Windows artefacts.
+
+### Fixed
+
+- Intel macOS release builds now use the correct hosted runner label in the release workflow.
+- Release packaging and GitHub Release artefact publication now generate the expected target-specific archives and clean up stale assets.

Cargo.lock

@@ -6,7 +6,7 @@ members = [
 resolver = "2"
 
 [workspace.package]
-version = "0.1.0"
+version = "0.1.1"
 edition = "2024"
 license = "Apache-2.0"
 

Cargo.toml

@@ -23,32 +23,35 @@ bitloops-inference describe-profile --config config.toml --profile openai_fast
 
 ## Config
 
-Profiles are defined under `[inference.profiles.<name>]`.
+`bitloops-inference` reads the Bitloops daemon inference config. Text-generation profiles live under `[inference.profiles.<name>]` and reference a runtime from `[inference.runtimes.<name>]`.
 
 ```toml
+[inference.runtimes.bitloops_inference]
+request_timeout_secs = 60
+
 [inference.profiles.openai_fast]
-kind = "openai_chat_completions"
-provider_name = "openai"
+task = "text_generation"
+driver = "openai_chat_completions"
+runtime = "bitloops_inference"
 model = "gpt-4.1-mini"
 base_url = "https://api.openai.com/v1/chat/completions"
 api_key = "${OPENAI_API_KEY}"
-temperature = 0.1
-timeout_secs = 60
+temperature = "0.1"
 max_output_tokens = 200
 
 [inference.profiles.ollama_local]
-kind = "ollama_chat"
-provider_name = "ollama"
+task = "text_generation"
+driver = "ollama_chat"
+runtime = "bitloops_inference"
 model = "qwen2.5-coder:14b"
 base_url = "http://127.0.0.1:11434/api/chat"
-temperature = 0.1
-timeout_secs = 120
+temperature = "0.1"
 max_output_tokens = 200
 ```
 
-String fields support `${ENV_VAR}` interpolation. Missing environment variables fail validation immediately.
+String fields support `${ENV_VAR}` interpolation. Missing environment variables fail validation immediately. Non-text-generation profiles in the same daemon config are ignored by `bitloops-inference`.
 
-## Supported provider kinds
+## Supported drivers
 
 - `openai_chat_completions`
 - `ollama_chat`
@@ -83,19 +86,19 @@ Example responses:
 Run config validation first:
 
 ```bash
-cargo run -p bitloops-inference -- validate-config --config ./config.toml
+cargo run -p bitloops-inference -- validate-config --config ./bitloops-daemon-config.toml
 ```
 
 Describe a profile:
 
 ```bash
-cargo run -p bitloops-inference -- describe-profile --config ./config.toml --profile ollama_local
+cargo run -p bitloops-inference -- describe-profile --config ./bitloops-daemon-config.toml --profile ollama_local
 ```
 
 Start the stdio runtime:
 
 ```bash
-cargo run -p bitloops-inference -- run --config ./config.toml --profile ollama_local
+cargo run -p bitloops-inference -- run --config ./bitloops-daemon-config.toml --profile ollama_local
 ```
 
 You can then write protocol lines to `stdin` manually or from another process.