xonecas.io
62159
Sy

symb / commit 99799e2

feat: show more provider examples

xonecas2026-02-16 99799e29af299e6bd258eb0d609a757af041e5cc parent eb64e8f

6 files changed

config.toml+16−1
@@ -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-codex"
4+default_provider = "zen-gemini"
55
66 # Ollama providers (local)
77 [providers.ollama-qwen]
@@ -40,6 +40,21 @@ endpoint = "https://opencode.ai/zen/v1"
4040 model = "claude-sonnet-4-5"
4141 api_key_name = "opencode_zen"
4242
43+[providers.zen-kimi]
44+endpoint = "https://opencode.ai/zen/v1"
45+model = "kimi-k2.5"
46+api_key_name = "opencode_zen"
47+
48+[providers.zen-glm]
49+endpoint = "https://opencode.ai/zen/v1"
50+model = "glm-4.7"
51+api_key_name = "opencode_zen"
52+
53+[providers.zen-gemini]
54+endpoint = "https://opencode.ai/zen/v1"
55+model = "gemini-3-pro"
56+api_key_name = "opencode_zen"
57+
4358 # Local vLLM (OpenAI-compatible)
4459 [providers.vllm-nanbeige]
4560 endpoint = "http://localhost:8000/v1"
docs/DESIGN.md+21−6
@@ -184,8 +184,29 @@ Make a generic modal with inputbox + list combo for re-use. The use that to make
184184 - Need to consider if this makes sense, or if the LLM should also be aware of user changes,
185185 and we should always send the diff for the LLM to apply?
186186
187+### Sub-Agent Tool
188+
189+Spawn a child LLM turn scoped to a single task. Useful for parallel work
190+or decomposing complex operations.
191+
192+- Have a short max iteration
193+- Customized prompt for single purpose agents
194+- TODO: Custom agents from community configs
195+
187196 ## Features waiting implementation for current version:
188197
198+### Community configs support:
199+
200+- CLAUDE/AGENTS/CURSOR/ETC.md (or just AGENTS.md at this point?) -- Is there a standard? need to ask an LLM that is not codex, otherwise it just tells me to do it.
201+- Custom subagent prompts for opencode/others custom agents
202+- Commands and skills implementation to match community expectations
203+
204+### Agent input interactivity:
205+
206+Use a special key `@` to spawn an autocomplete. This can match files, skills, commands or subagents.
207+When the user selects a match with `Enter` it replaces the @ with the selected item once the message is sent to the
208+LLM
209+
189210 Testing current feature set
190211
191212 # TODO (after features)
@@ -205,12 +226,6 @@ Testing current feature set
205226 Pause before executing tool calls. Show tool name + args in a dialog. User
206227 approves/rejects. Configurable per-tool permissions in `config.toml` (allow,
207228 ask, deny). Some tools (Read/Grep) default allow, mutations (Edit) default ask.
208-
209-### Sub-Agent Tool
210-
211-Spawn a child LLM turn scoped to a single task. Useful for parallel work
212-or decomposing complex operations.
213-
214229 ### Parallel Tool Execution
215230
216231 Execute multiple independent tool calls concurrently within a single LLM
internal/mcptools/subagent.go+1−1
@@ -37,7 +37,7 @@ type SubAgentArgs struct {
3737 // NewSubAgentTool creates the SubAgent tool definition.
3838 func NewSubAgentTool() mcp.Tool {
3939 return mcp.Tool{
40- Name: "SubAgent",
40+ Name: "SubAgent",
4141 Description: `Spawn a sub-agent to handle a focused task. The sub-agent runs with the same tools but cannot spawn further sub-agents. Use this to decompose complex tasks into smaller, manageable pieces. The sub-agent's work is returned as a summary.`,
4242 InputSchema: json.RawMessage(`{
4343 "type": "object",
internal/provider/opencode.go+8−7
@@ -3,7 +3,6 @@ package provider
33 import (
44 "context"
55 "encoding/json"
6- "fmt"
76 "net/http"
87 "strings"
98
@@ -39,7 +38,10 @@ const (
3938
4039 // opencodeModelEndpoints overrides the prefix-based fallback in opencodeEndpointForModel.
4140 // Only list models whose endpoint differs from what the prefix logic would choose.
42-var opencodeModelEndpoints = map[string]string{}
41+var opencodeModelEndpoints = map[string]string{
42+ "gemini-3-pro": "/models/gemini-3-pro",
43+ "gemini-3-flash": "/models/gemini-3-flash",
44+}
4345
4446 // NewOpenCode creates a new OpenCode Zen provider.
4547 func NewOpenCode(endpoint, model, apiKey string) *OpenCodeProvider {
@@ -71,17 +73,16 @@ func (p *OpenCodeProvider) ChatStream(ctx context.Context, messages []Message, t
7173 switch endpoint {
7274 case opencodeMessagesEndpoint:
7375 return p.chatStreamAnthropic(ctx, messages, tools)
74- case opencodeChatCompletionsEndpoint:
75- return p.chatStreamOpenAI(ctx, messages, tools)
7676 case opencodeResponsesEndpoint:
7777 return p.chatStreamResponses(ctx, messages, tools)
7878 default:
79- return nil, fmt.Errorf("opencode model %q uses unsupported endpoint %q", p.model, endpoint)
79+ // OpenAI-compatible chat completions (also used for Gemini, GLM, Kimi, etc.)
80+ return p.chatStreamOpenAI(ctx, endpoint, messages, tools)
8081 }
8182 }
8283
8384 // chatStreamOpenAI streams via the OpenAI-compatible /chat/completions endpoint.
84-func (p *OpenCodeProvider) chatStreamOpenAI(ctx context.Context, messages []Message, tools []Tool) (<-chan StreamEvent, error) {
85+func (p *OpenCodeProvider) chatStreamOpenAI(ctx context.Context, endpoint string, messages []Message, tools []Tool) (<-chan StreamEvent, error) {
8586 openaiTools := toOpenAITools(tools)
8687
8788 customReq := openCodeRequest{
@@ -99,7 +100,7 @@ func (p *OpenCodeProvider) chatStreamOpenAI(ctx context.Context, messages []Mess
99100
100101 reader, err := httpDoSSE(ctx, httpRequestConfig{
101102 client: p.httpClient,
102- url: p.baseURL + opencodeChatCompletionsEndpoint,
103+ url: p.baseURL + endpoint,
103104 body: body,
104105 headers: p.authHeaders(),
105106 provider: p.name,
internal/tui/messages.go+0−1
@@ -161,7 +161,6 @@ func (m Model) sendDiffToLLM() tea.Cmd {
161161 }
162162 }
163163
164-
165164 func (m Model) waitForLLMUpdate() tea.Cmd {
166165 ch := m.updateChan
167166 return func() tea.Msg {
internal/tui/update.go+0−1
@@ -17,7 +17,6 @@ import (
1717 "github.com/xonecas/symb/internal/tui/modal"
1818 )
1919
20-
2120 // ---------------------------------------------------------------------------
2221 // Update
2322 // ---------------------------------------------------------------------------