|
13 | 13 |
|
14 | 14 | Code Web Chat (CWC) is a chatbot-first AI coding tool. **Start in VS Code (Cursor, Antigravity, VSCodium, etc.), then continue in ChatGPT, Gemini, Claude (plus 15+ more)** or get the job done without leaving your editor! |
15 | 15 |
|
16 | | -It's free, fully-featured and doesn't collect any telemetry. |
| 16 | +It's free, fully-featured and doesn't collect telemetry. |
17 | 17 |
|
18 | 18 | Get involved! Join our [discord server](https://discord.gg/KJySXsrSX5). |
19 | 19 |
|
@@ -47,7 +47,7 @@ CWC flips this! |
47 | 47 | Here, selected in the explorer view files are sent to the model with your instructions so it has everything it needs to do the task. Get accurate multi-file edits in record time! |
48 | 48 |
|
49 | 49 | > [!TIP] |
50 | | -> **Not sure what files to select?** Search files using phrase, keywords and **natural language**. |
| 50 | +> **Not sure what files to select?** You can search files using phrase, keywords or **natural language** globally or within the selected folder. |
51 | 51 |
|
52 | 52 | **Generated prompts are structured as follows:** |
53 | 53 |
|
@@ -116,17 +116,6 @@ Find a complete set of relevant files according to the following query: |
116 | 116 |
|
117 | 117 | </details> |
118 | 118 |
|
119 | | -## Prompt caching |
120 | | - |
121 | | -With first-class support for [prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching), CWC enables unmatched cost-efficiency and speed. |
122 | | - |
123 | | -- Preserve your chatbot quota for longer |
124 | | -- Lower input token costs by up to 90% |
125 | | -- Reduce latency by up to 80% |
126 | | - |
127 | | -**How it works?** \ |
128 | | -Selected files are ordered by modification and selection recency, and because task instructions are always placed at the very end of the prompt, it's submitted largery unchanged across tasks! |
129 | | - |
130 | 119 | ## Enabling autofill |
131 | 120 |
|
132 | 121 | Install the [browser extension](https://github.com/robertpiosik/CodeWebChat/tree/dev/apps/browser) and never copy and paste again. |
@@ -163,24 +152,15 @@ Install the [browser extension](https://github.com/robertpiosik/CodeWebChat/tree |
163 | 152 | > [!NOTE] |
164 | 153 | > Use [forwarding](https://code.visualstudio.com/docs/debugtest/port-forwarding) of port _55155_ when using remote machine via SSH. |
165 | 154 |
|
166 | | -## Calling APIs |
167 | | - |
168 | | -All the essential API tools to get the job done. Bring Your Own Key (BYOK) for a model provider of choice. Use self-hosted model provider like [Ollama](https://ollama.com/search) or [LM Studio](https://lmstudio.ai/). |
| 155 | +## Prompt caching |
169 | 156 |
|
170 | | -- **Intelligent Update**—Integrate changes from malformed responses. |
171 | | -- **Edit Files**—Pair-programming using natural language instructions. |
172 | | -- **Code at Cursor**—Accurate inline code from reasoning models. |
173 | | -- **Find Relevant Files**—Find files using natural language description. |
174 | | -- **Commit Messages**—Meaningful summaries of changes in your style. |
175 | | -- **Voice Input**—Transcribe speech to text in the prompt field. |
| 157 | +CWC orders selected files by modification recency. This, combined with its default single-turn interactions, creates a workflow that is extremely cache-friendly. Make your chatbot quota last longer, lower input token costs by up to 90%, and reduce latency by up to 80%. |
176 | 158 |
|
177 | 159 | ## Privacy |
178 | 160 |
|
179 | | -The extension doesn’t communicate with any server for so-called telemetry. |
180 | | - |
181 | | -The editor-browser communication for chatbot autofill runs on local WebSockets. |
182 | | - |
183 | | -When calling APIs, requests are sent directly to the configured model providers. |
| 161 | +- zero telemetry |
| 162 | +- browser communication over localhost |
| 163 | +- model providers called directly |
184 | 164 |
|
185 | 165 | ## Commands |
186 | 166 |
|
|
0 commit comments