1 # Changelog
2
3 ## Unreleased
4
5 ### Fixed
6
7 * **Qwen3-14B on-device inference** (#5): selecting Qwen3-14B and prompting it failed with
8 `Inference failed: inference error: Invalid sampling probability at index 0: NaN. The model
9 likely produced NaN/Inf logits.` This is a half-precision (F16) overflow in the `onde`
10 inference engine on GPUs without bfloat16 support (notably Intel Macs); the 14B model's
11 attention scores exceed F16's range and become `NaN`. Fixed in the engine by preferring F32
12 over the F16 fallback (see `onde/patches/`), surfaced here once the `onde` dependency is
13 bumped. siGit also now shows an actionable hint instead of the raw engine error.
14
15 ## 1.2.1
16
17 Stabilizes the Zed/ACP integration and finishes the cloud-tier wiring on top of 1.2.0.
18
19 ### What changed
20
21 - Fixed a Zed crash by keeping model-picker labels ASCII in the ACP model selector
22 - Wired ACP auth, cloud tiers, and slash commands into the Zed panel, including a `/reload` command to re-sync session state in place
23 - Fixed the TUI so the loaded-model checkmark appears once a download completes
24 - Synced bundled agent skills with the current code and added `CLAUDE.md`
25
26 ## 1.2.0
27
28 Adds siGit Code Cloud — a hosted inference tier alongside on-device models.
29
30 ### What changed
31
32 - Added siGit Code Cloud with cloud-tier routing, pointed at `sigit.si`
33 - Added account management slash commands and surfaced cloud tiers in `/models`
34 - Carried forward from 1.1.0: ACP SDK v0.13, refreshed dependencies and branding
35
36 ## 1.1.0
37
38 Bumps the ACP SDK to v0.13 and pulls in updated dependencies.
39
40 ### What changed
41
42 - Updated `agent-client-protocol` from v0.11 to v0.13
43 - Updated `onde` to 1.1.2
44 - Refreshed branding and skill metadata
45
46 ## 1.0.4
47
48 This release tightens up the terminal experience and finishes a few release-facing cleanup items.
49
50 ### What changed
51
52 - Added bold rich-text rendering in the TUI for assistant replies, so `**text**` now displays with terminal styling instead of raw markdown markers
53 - Refreshed the bundled skill metadata to follow the current Agent Skills `SKILL.md` format
54 - Synced the crate release metadata for the `1.0.4` cut
55
56 ## 1.0.3
57
58 This is the cleanup release for the editor-side startup problems.
59
60 ### What changed
61
62 - Fixed ACP sessions failing on the first real prompt because the server claimed the model was ready before anything had actually been loaded
63 - Changed ACP startup so the default model loads lazily on the first non-slash prompt instead of pretending it is already in memory
64 - Kept `initialize` and `session/new` lightweight while still sending proper progress updates once model loading begins
65 - Updated the Onde integration to `1.0.0`
66 - Removed a few dependencies we were no longer using
67
68 ## 1.0.2
69
70 This release was supposed to fix the ACP auth breakage. It did fix the stdout pollution problem, but it turned out not to be the whole story.
71
72 ### What changed
73
74 - Delayed model loading in ACP mode so startup diagnostics would not leak into protocol stdout during the auth handshake
75 - Tightened up the ACP startup path for editor integrations
76 - Refreshed some README wording while cutting `1.0.2`
77
78 ## 1.0.1
79
80 The first patch after `1.0.0` was mostly about making model loading and model switching feel less opaque.
81
82 ### What changed
83
84 - Added ToolCall-based progress UI for startup model loading and downloading
85 - Improved model-switch progress reporting in ACP clients
86 - Fixed model-load error handling so failed switches did not leave the UI in a weird state
87 - Cleaned up a few status messages and docs while the release was going out
88
89 ## 1.0.0 (2026)
90
91 siGit Code has been living in real smbCloud repos for a while now. At some point it stopped feeling like an experiment, so we called it 1.0.
92
93 ### What this release is
94
95 siGit Code is a local coding agent. It runs a quantized model on your machine, talks to editors over ACP, and can read files, run commands, fetch web pages, and write code without sending your project to a hosted API.
96
97 You can install it with Cargo, pip, npm, or Homebrew and use it like any other tool on your machine.
98
99 ### What shipped in 1.0
100
101 #### Editor integration
102
103 This is the core of the project.
104
105 Zed and VS Code can talk to siGit Code over ACP. Multi-turn sessions work. Tool calling works. Session forking works. Working-directory context works. That was the original goal, and it feels solid now.
106
107 #### Terminal UI
108
109 The terminal UI started as a side quest and turned out to be useful. You get a full-screen ratatui chat, streaming tokens, a spinner while the model is busy, and a model picker you can open in the middle of a session.
110
111 It runs on macOS and Linux. Windows gets ACP and editor mode for now. The Windows terminal UI is still unfinished.
112
113 #### Tool calling
114
115 This is the part that makes siGit Code feel like an agent instead of a chat box. The loop can run up to 10 rounds per message.
116
117 Available tools:
118
119 - `read_file` / `write_file` / `delete_file`
120 - `list_directory` / `search_files`
121 - `run_command`, with an optional working directory
122 - `read_website`, which fetches a URL and strips it down to readable text
123
124 The model can call a tool, inspect the result, and keep going until it has a real answer.
125
126 #### Model support
127
128 The model list ended up wider than we expected for 1.0:
129
130 - Qwen 3 1.7B, 4B, 8B, and 14B
131 - Qwen 2.5 1.5B and 3B
132 - Qwen 2.5 Coder 1.5B, 3B, and 7B
133 - DeepSeek Coder 6.7B
134
135 They are all GGUF models and they all come from Hugging Face on first run.
136
137 Qwen 3 is the interesting one. It uses extended thinking mode. The model reasons inside `<think>...</think>` blocks before answering. The TUI strips those blocks out and renders them dimmed above the reply, so you can see what happened without turning the whole conversation into noise.
138
139 The 8B model is the desktop default. Mobile stays on 1.7B because iOS gives apps roughly 2 to 3 GB of memory, and we learned the hard way that 3B can blow up on an iPhone 16e.
140
141 #### Model picker
142
143 The model picker shows:
144
145 - what is already cached locally
146 - what can be downloaded
147 - which models support tool calling
148 - whether the local cache looks healthy
149
150 You can open it with `/models` in the TUI or through the editor config option. Switching models happens in the background and the UI stays alive while the download or load is in progress.
151
152 #### smbCloud context
153
154 siGit Code knows smbCloud repos better than a generic coding assistant does. It understands the difference between platform-user flows and tenant-app auth flows, how `Project`, `FrontendApp`, `AuthApp`, and GresIQ fit together, and why Next.js SSR deploys are not the same thing as the generic git-push path.
155
156 Outside smbCloud, it backs off and behaves like a normal coding agent.
157
158 #### Distribution
159
160 Distribution took an unreasonable amount of time, honestly.
161
162 There are prebuilt binaries for macOS, Linux, and Windows on both arm64 and x64 where relevant, plus install paths through Cargo, PyPI, npm, and Homebrew:
163
164 - `cargo install sigit`
165 - `pip install sigit-code`
166 - `npm install -g @smbcloud/sigit`
167 - `brew install sigit`
168
169 Getting that whole pipeline to behave across CI, crates.io, PyPI, npm, and Homebrew was basically its own project.
170
171 ### What still does not work
172
173 The Windows terminal UI is still missing.
174
175 ACP and editor mode work on Windows. The part that is still missing is the interactive full-screen terminal UI. The blocker is Unix-specific terminal handling that has not been abstracted cleanly yet.
176
177 ### Changes since 0.1.2
178
179 - Added Qwen 3 14B support
180 - Added Qwen 3 `<think>` block parsing and separate rendering in the TUI
181 - Moved all TUI code into `#[cfg(unix)]`, which fixed a pile of dead-code errors on Windows CI
182 - Added live download progress during model switches, including cancellation with Ctrl+C
183 - Added model download and loading progress in the Zed agent config panel
184 - Added an animated spinner during model switching
185 - Added Qwen 2.5 Coder 7B
186 - Added downloadable models to the picker, not just locally cached ones
187 - Made model selection persist across restarts
188 - Added session working-directory support
189 - Moved model picker logic into a platform-independent module so Windows can compile without the TUI
190 - Added the `/models N` shortcut for picking a model by number
191 - Added the `read_website` tool
192 - Improved `read_file` handling and empty-reply detection
193 - Added async tool execution
194 - Fixed CI cross-compilation for macOS, iOS, Linux, and Windows
195 - Added npm, PyPI, and Homebrew distribution
196
197 ---
198
199 *© 2026 [smbCloud](https://smbcloud.xyz/) (Splitfire AB).*