Repository navigation
Paddler 5.0: completions, embeddings and decisions from one loaded model, kev decisions and TypeSafe compatibility - #270
Draft
mcharytoniuk wants to merge 129 commits into
Draft
mcharytoniuk wants to merge 129 commits into
mcharytoniuk wants to merge 129 commits into
Conversation
…ly and through a TypeSafe-compatible API
…t only against real Paddler instances
…processes that exit before they are ready
… desired state PUT once the balancer shuts down
…and servers with real ones or direct checks of the real handlers
… Qwen3.8 decisions
…d escape control-token lookalikes the way kev does
… TypeSafe floats exactly like kev
…he issues of a replaced desired state
…every state change a single owner, cancel replaced downloads, detect dead peers, and make the Makefile portable
…ser cache, and test startup from the stored state, PUT ordering, untouched missing state files, and stuck pipelines
…tead of hand-rolled panics in tests, and share fixtures for agent controllers and generated text
… so multimodal prompts are tokenized only once
…em One requests, keep text.format schemas as raw JSON, compute OpenAI failure statuses once, and subscribe to agent availability before the first scan
…s, and test that new desired states preempt retries and that unread clients release agent slots
…trust the agent's softmax instead of normalizing again
…dings 0.18 so each chunk is checked once
… keep distinct embedding contents in a named struct, and expect Responses usage and text instead of panicking by hand
…k is classified once and its ingestion cannot fail
…late and assert exact dispatch per inference mode
…r seams and test that each refuses the generation
…eachable management address, and list the suites the Makefile actually has in the test skills
…t call them, build test generations through the production preparer, and cap batch sizes with BatchSize::at_most
…alancers that vanish under an agent socket, route a decision refusal through Cluster::decide, and serialize forged notifications in the tests that send them
…tream error codes through a tested harness helper
…nnot fail, and send error responses through the session's logging send
…ing scheduler can reach
…gone at tool call resolution, a deferred prompt, and empty or unparsable tool call bodies
…rocess can record its coverage, and advance generations with logging enabled as production does
…n that excludes every candidate
…ut instead of a sampler failure
…serializer, and media marker with failing stand-ins
…ich only test doubles implemented
…nrecognized format
…as an unrecognized format
…Code and embedded static files in coverage
…NT, and report a Hugging Face client that cannot start
…request, dropping errors that cannot occur
… bodies, a startup-built model resolver and llama-cpp-bindings 0.23
…, and drop dependencies no crate uses
…ambiguous names, and drop leftover dependency declarations
…end TypeSafe a rejection the newer SDK still forwards
…nd adopt the React hooks rules it brings
…b 1, minijinja 3, reqwest 0.13 and the other new APIs
…e pre-cancelled resolution test after what it checks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
One inference mode per cluster, kev decisions and TypeSafe compatibility.