tui

Documentation

Overview

Package tui implements the interactive terminal UI shown when cc-fleet is run bare (no subcommand) from an interactive terminal. It is a thin arrow-key front end over the same internal packages the subcommands use (userops for provider CRUD, teardown for teammate discovery) so the two never drift. It is gated behind a tty check in cmd/cc-fleet so pipes, CI, and --json callers never block on the bubbletea event loop.

Index

• Variables
• func Run(verbose bool) error
• type Model
  • func NewModel() Model
  • func (m Model) Init() tea.Cmd
  • func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd)
  • func (m Model) View() string
• type OAITemplate
• type Template

Variables

var OAITemplates = []OAITemplate{
	{
		Label:       "OpenAI · Responses API (official)",
		Name:        "openai",
		Protocol:    config.ProtocolOpenAIResponses,
		UpstreamURL: "https://api.openai.com/v1",
		Note:        "High-fidelity reasoning surface; pick the default model from the probed list.",
	},
	{
		Label:       "OpenAI · Chat Completions (official)",
		Name:        "openai-chat",
		Protocol:    config.ProtocolOpenAIChat,
		UpstreamURL: "https://api.openai.com/v1",
	},
	{
		Label:    "OpenAI-compatible · Chat (Groq / Together / Fireworks / vLLM …)",
		Name:     "",
		Protocol: config.ProtocolOpenAIChat,
		Note:     "Set upstream_url to the provider's base (often …/v1, e.g. https://api.groq.com/openai/v1).",
	},
}

OAITemplates seeds the OpenAI-protocol picker. A synthetic "Custom" entry is appended by the picker, so it is intentionally NOT in this slice.

var Templates = []Template{

	{
		Name:           "deepseek",
		Label:          "DeepSeek",
		BaseURL:        "https://api.deepseek.com/anthropic",
		ModelsEndpoint: "https://api.deepseek.com/v1/models",
		DefaultModel:   "deepseek-v4-flash",
	},
	{
		Name:           "kimi",
		Label:          "Moonshot Kimi",
		BaseURL:        "https://api.moonshot.cn/anthropic",
		ModelsEndpoint: "https://api.moonshot.cn/anthropic/v1/models",
		DefaultModel:   "kimi-latest",
	},
	{
		Name:           "glm",
		Label:          "Zhipu GLM",
		BaseURL:        "https://open.bigmodel.cn/api/anthropic",
		ModelsEndpoint: "https://open.bigmodel.cn/api/paas/v4/models",
		DefaultModel:   "glm-4.6",
	},
	{
		Name:           "qwen",
		Label:          "Qwen (Alibaba DashScope)",
		BaseURL:        "https://dashscope.aliyuncs.com/apps/anthropic",
		ModelsEndpoint: "https://dashscope.aliyuncs.com/compatible-mode/v1/models",
		DefaultModel:   "qwen-max",
		Note:           "DashScope endpoints vary by region and plan; if the probe fails on add, correct base_url per the DashScope docs.",
	},
	{
		Name:           "minimax",
		Label:          "MiniMax",
		BaseURL:        "https://api.minimaxi.com/anthropic",
		ModelsEndpoint: "https://api.minimaxi.com/v1/models",
		DefaultModel:   "MiniMax-M2",
	},
	{
		Name:           "xiaomimimo",
		Label:          "Xiaomi MiMo",
		BaseURL:        "https://api.xiaomimimo.com/anthropic",
		ModelsEndpoint: "https://api.xiaomimimo.com/v1/models",
		DefaultModel:   "mimo-v2.5-pro",
		Note:           "models_endpoint sits at the host root (not under /anthropic); probe-verified.",
	},

	{
		Name:           "zai",
		Label:          "Zhipu GLM (z.ai, international)",
		BaseURL:        "https://api.z.ai/api/anthropic",
		ModelsEndpoint: "https://api.z.ai/api/paas/v4/models",
		DefaultModel:   "glm-4.6",
		Note:           "International GLM site; same shape as the mainland endpoint but a separate API key.",
	},
	{
		Name:           "stepfun",
		Label:          "StepFun",
		BaseURL:        "https://api.stepfun.com/step_plan",
		ModelsEndpoint: "https://api.stepfun.com/v1/models",
		DefaultModel:   "step-3.5-flash-2603",
		Note:           "base_url is the step-plan (coding) endpoint; models_endpoint (/v1) is inferred — the probe on add verifies it.",
	},
	{
		Name:           "longcat",
		Label:          "LongCat (Meituan)",
		BaseURL:        "https://api.longcat.chat/anthropic",
		ModelsEndpoint: "https://api.longcat.chat/v1/models",
		DefaultModel:   "LongCat-Flash-Chat",
		Note:           "models_endpoint is inferred — the probe on add verifies it.",
	},
	{
		Name:           "volcengine",
		Label:          "Volcengine Ark (ByteDance)",
		BaseURL:        "https://ark.cn-beijing.volces.com/api/coding",
		ModelsEndpoint: "https://ark.cn-beijing.volces.com/api/v3/models",
		DefaultModel:   "ark-code-latest",
		Note:           "Coding-plan endpoint with a fixed code model; the general API addresses models by endpoint id (ep-...), so the model list may come back empty — enter the model by hand.",
	},
	{
		Name:           "doubao",
		Label:          "Doubao Seed (ByteDance Volcengine)",
		BaseURL:        "https://ark.cn-beijing.volces.com/api/compatible",
		ModelsEndpoint: "https://ark.cn-beijing.volces.com/api/v3/models",
		DefaultModel:   "doubao-seed-2-0-code-preview-latest",
		Note:           "Same endpoint-id scheme as Volcengine Ark; confirm the model name/list in the Ark console.",
	},
	{
		Name:           "qianfan",
		Label:          "Baidu Qianfan",
		BaseURL:        "https://qianfan.baidubce.com/anthropic/coding",
		ModelsEndpoint: "https://qianfan.baidubce.com/v2/models",
		DefaultModel:   "qianfan-code-latest",
		Note:           "Coding-plan endpoint with a fixed code model; confirm the model list / general model in the Qianfan docs.",
	},
	{
		Name:           "bailing",
		Label:          "Ant Ling (Bailing)",
		BaseURL:        "https://api.tbox.cn/api/anthropic",
		ModelsEndpoint: "https://api.tbox.cn/v1/models",
		DefaultModel:   "Ling-2.5-1T",
		Note:           "models_endpoint is inferred — the probe on add verifies it.",
	},
}

Templates is the built-in seed table. Order is the display order in the add wizard's picker. A synthetic "Custom" entry (empty fields) is appended by the picker itself, so it is intentionally NOT in this slice.

Functions

func Run

func Run(verbose bool) error

Run starts the bubbletea program against stdin/stdout. The caller is responsible for ensuring those are a terminal (see cmd/cc-fleet/tui.go). verbose routes the session's step trace to a per-pid 0600 log file.

Types

type Model

type Model struct {
	// contains filtered or unexported fields
}

Model is the root bubbletea model.

func NewModel

func NewModel() Model

NewModel returns the initial model. It normally parks on the Model Providers list (the hub) with loading=true so Init can kick off the provider load. On a first run where agent-teams looks unconfigured (and the user hasn't dismissed the nudge), it instead opens on the agent-teams setup screen; the hub loads when the user leaves setup via toList.

NewModel is only ever called from tui.Run, which cmd/cc-fleet gates to the bare-interactive both-TTY path — so the onboarding probe here never runs for spawn/subagent/piped/agent callers.

func (Model) Init

func (m Model) Init() tea.Cmd

Init satisfies tea.Model: load the provider list so the home screen is populated as soon as the program starts. On the setup screen there's nothing to load yet — toList kicks off loadProviders when the user proceeds.

func (Model) Update

func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd)

Update is the single tea.Model entry point. Async results (providersMsg etc.) are handled regardless of screen unless they implement screenOwnedAsyncMsg — in that case Update drops the message when the user has navigated off the owning screen, so e.g. a slow models-fetch result can't reach the provider list after the user esc'd back. Key handling dispatches on the active screen. ctrl+c always quits.

func (Model) View

func (m Model) View() string

View satisfies tea.Model.

type OAITemplate

type OAITemplate struct {
	Label        string // picker label
	Name         string // default provider id
	Protocol     string // config.ProtocolOpenAIChat | config.ProtocolOpenAIResponses
	UpstreamURL  string // real OpenAI-compatible base, usually ending in /v1
	DefaultModel string // suggested default (blank → pick from the probed list)
	Note         string
}

OAITemplate seeds the OpenAI-protocol add form. base_url is the loopback conversion-daemon URL (auto-assigned on add), so it is NOT a template field; upstream_url is the real OpenAI-compatible endpoint and Protocol selects the wire surface (Chat Completions vs Responses).

type Template

type Template struct {
	Name           string // providers.toml table name (lowercase id)
	Label          string // display name in the picker
	BaseURL        string // ANTHROPIC_BASE_URL
	ModelsEndpoint string // /v1/models URL used for the probe + refresh
	DefaultModel   string // suggested default model id
	Note           string // optional caveat shown in the picker preview
}

Template is a built-in provider seed: prefill values for the add wizard so a user picking "DeepSeek" doesn't have to type the base_url / models_endpoint by hand. These are SEEDS — provider URLs and model ids drift over time, so the add wizard lets the user confirm or edit every field, and `cc-fleet add` still probes /v1/models before committing.

Each entry follows the provider's published Anthropic-compatible endpoint. An entry WITHOUT a Note has a models_endpoint that has been probe-verified; an entry WITH a Note has an inferred or plan-specific endpoint flagged for the user to confirm (the probe on add does this automatically). Only first-party LLM providers that run their own models are seeded — not Claude-relay aggregators, which run Claude and defeat the point of a provider teammate.