xonecas.io
62159
Sy

symb / commit 82b7fff

chore: docs and config cleanup

xonecas2026-02-16 82b7fff90161f7ff81abdc90804c0b1053d96e51 parent 7c18cf9

2 files changed

config.toml+8−17
@@ -1,7 +1,7 @@
11 # Symb Configuration File
22
33 # Default provider (optional - if not set, first provider in map is used)
4-default_provider = "zen-opus"
4+default_provider = "zen-codex"
55
66 # Ollama providers (local)
77 [providers.ollama-qwen]
@@ -14,21 +14,6 @@ endpoint = "http://localhost:11434"
1414 model = "qwen3:4b"
1515 temperature = 0.2
1616
17-[providers.ollama-llama]
18-endpoint = "http://localhost:11434"
19-model = "llama3.1:8b"
20-temperature = 0.2
21-
22-[providers.ollama-llama32]
23-endpoint = "http://localhost:11434"
24-model = "llama3.2"
25-temperature = 0.2
26-
27-[providers.mfd-deepseek]
28-endpoint = "http://localhost:11434"
29-model = "MFDoom/deepseek-r1-tool-calling:14b"
30-temperature = 0.2
31-
3217 # OpenCode Zen providers (cloud)
3318 [providers.zen-nano]
3419 endpoint = "https://opencode.ai/zen/v1"
@@ -42,6 +27,12 @@ model = "big-pickle"
4227 api_key_name = "opencode_zen"
4328 temperature = 0.2
4429
30+[providers.zen-codex]
31+endpoint = "https://opencode.ai/zen/v1"
32+model = "gpt-5.2-codex"
33+api_key_name = "opencode_zen"
34+temperature = 0.2
35+
4536 [providers.zen-opus]
4637 endpoint = "https://opencode.ai/zen/v1"
4738 model = "claude-opus-4-6"
@@ -64,4 +55,4 @@ ttl_hours = 24
6455
6556 [git]
6657 # TODO: When adding worktrees later
67-enable_sandbox = false
58+# enable_sandbox = false
docs/DESIGN.md+5−4
@@ -119,12 +119,9 @@ Needs an undo.
119119 <time elapsed> <timestamp> <toekns in current count>/<tokens out current count> (<total tokens in context at current count).
120120 <undo if most recent agent message>
121121 ```
122-
123-Refactor so it shows two lines as above with text right aligned. Remove existing `---` separator completly
124122 **DONE**
125123
126124 - Read tool caps output at 500 lines / 20k chars; tells LLM to use line ranges for larger files.
127-- Old tool results are compacted to their header line after the assistant responds, preventing resend of full payloads on subsequent rounds.
128125
129126 2. Gather baseline data with no context management. Then study:
130127 **INPROGRESS**
@@ -186,7 +183,11 @@ Testing current feature set
186183 ### Code viewer pane -> Becomes Code editor pane
187184
188185 - Enable the same editting abilities as the agent input box.
189-- <ctrl + s> Saves the file. (Do we just save it, or we send a message to the LLM to apply the edit?)
186+- <ctrl + s> Saves the file:
187+ - File has no changes aside from the user's changes -> save file. Fast, simple.
188+ - File has changes coliding with the user's changes -> send to llm to merge
189+ - Need to consider if this makes sense, or if the LLM should also be aware of user changes,
190+ and we should always send the diff for the LLM to apply?
190191
191192 # TODO (after features)