DeepSeek Harness 0.2.2-rc.4 fixes LiveShare and handles unexpected model responses better
TL;DR A release candidate that fixes LiveShare session creation in the desktop app and handles out-of-spec responses from the model more gracefully.
In plain words: a “harness” is the program that wraps an AI model so it can act as an agent (read files, run tools, keep a session). This update fixes a sharing feature and makes the harness more forgiving when the model answers in a format it did not expect.
What changed
v0.2.2-rc.4 of the DeepSeek Harness fixes the creation of LiveShare sessions in the desktop app and improves how it handles unexpected responses from the model. It is still a release candidate, and LiveShare remains experimental.
Who is affected
Anyone on the 0.2.x line, and especially people who point the harness at a self-hosted model.
Why it matters
Local inference servers sometimes return a malformed tool call or a tool name that does not exist. In earlier versions that could end the agent’s turn instead of returning the error to the model so it can try again.
What to do
Try the RC in a test setup first; move to the final 0.2.2 when it ships.