util

func ComputeMessagesContextLength ¶ added in v0.7.0

func ComputeMessagesContextLength(ctx *lmcli.Context, conv *conversation.Conversation, messages []conversation.Message) []int

ComputeMessagesContextLength computes token counts for messages in a conversation. For system messages, the conversation's agent tools are included in the count. Returns a slice of token counts aligned with the messages slice.

func FormatForExternalPrompt ¶

func FormatForExternalPrompt(messages []conversation.Message, system bool) string

func LookupConversation ¶

func LookupConversation(ctx *lmcli.Context, shortName string) *conversation.Conversation

LookupConversation either returns the conversation found by the short name or exits the program

func LookupConversationE ¶

func LookupConversationE(ctx *lmcli.Context, shortName string) (*conversation.Conversation, error)

func RenderConversation ¶

func RenderConversation(ctx *lmcli.Context, messages []conversation.Message, spaceForResponse bool)

RenderConversation renders the given messages to TTY, with optional space for a subsequent message. spaceForResponse controls how many '\n' characters are printed immediately after the final message (1 if false, 2 if true). Message blocks are built in parallel for faster rendering.

func RenderMessage ¶

func RenderMessage(ctx *lmcli.Context, m *conversation.Message)

func RunCompletionLoop ¶ added in v0.7.0

func RunCompletionLoop(
	ctx context.Context,
	appCtx *lmcli.Context,
	sess *session.Session,
	showThinking bool,
	perm api.PermissionMode,
) error

RunCompletionLoop drives the agent completion + tool-call loop for CLI commands. The session must have its agent set before calling. It starts a completion step, renders chunks to stdout, evaluates tool calls against the permission mode, executes them (with user confirmation when needed), and continues until a response has no tool calls or all tools are rejected. Cancelling ctx stops the loop.

func RunSession ¶ added in v0.7.0

func RunSession(cmd *cobra.Command, ctx *lmcli.Context, sess *session.Session)

RunSession configures the session from the application context, renders existing messages, and runs the agent completion loop (streaming output to stdout with auto-approved tool calls).