Documentation
¶
Overview ¶
Package llm handles LLM-based generation tasks: prompt rendering, template embedding, runner invocation, and output parsing. Finders (read side) and handlers (write side) both consume this package — it owns the "call an LLM with a structured prompt" concern.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func ComputePromptHash ¶
ComputePromptHash returns the hex-encoded SHA-256 hash of the rendered prompt. Uses the first 16 bytes (32 hex chars) — enough for collision avoidance.
func FormatEntryForPrompt ¶
FormatEntryForPrompt formats an entry as readable text for inclusion in a prompt.
Types ¶
type LLMMetadata ¶
type LLMMetadata struct {
TotalCostUSD float64
InputTokens int
OutputTokens int
CacheReadTokens int
CacheCreateTokens int
NumTurns int
Duration time.Duration
DurationAPI time.Duration
Models map[string]ModelUsage
}
LLMMetadata holds agent-neutral per-call metrics from the LLM provider.
type ModelUsage ¶
type ModelUsage struct {
InputTokens int
OutputTokens int
CacheReadTokens int
CacheCreateTokens int
CostUSD float64
}
ModelUsage holds per-model token and cost metrics.
type PreflightResult ¶
type PreflightResult struct {
Findings []Finding
}
PreflightResult holds the parsed findings from a pre-flight validator run. An empty Findings slice means the validator reported no findings.
func Preflight ¶
func Preflight(ctx context.Context, runner Runner, entry *model.Entry, graph *model.Graph, configuredLanguage string) (*PreflightResult, error)
Preflight runs the pre-flight validator against the given entry and graph. Returns the parsed result regardless of finding severity. Returns an error only for infrastructure failures (runner error, template error, parse error).
Participant validation is handled by mechanical checks in the finders layer (see finders.mechanicalPreflight) — the LLM participant-drift rubric is retired per plan d-cpt-d34 AC 9.
configuredLanguage is the graph authoring language (locale code) from `.sdd/config.yaml` (empty string when unset — English default). It feeds the language-drift check which flags entries whose description prose does not match the configured language.
func (*PreflightResult) HasBlocking ¶
func (r *PreflightResult) HasBlocking() bool
HasBlocking reports whether any finding blocks entry creation. Currently only SeverityHigh blocks.
type Request ¶ added in v0.2.0
Request carries the two-part prompt submitted to a Runner. SystemPrompt holds the stable portion (instructions, structural rules) — providers that support prompt caching treat this as the cacheable prefix. UserPrompt holds the per-call variable portion (entry content, refs). Runners that can't distinguish system from user concatenate them with SystemPrompt first.
func RenderSummaryPrompt ¶
RenderSummaryPrompt renders the summary prompt for an entry. Returns a Request with the full rendered prompt in UserPrompt; the system/user split is introduced when templates are refactored (see the plan decision).
func (Request) Combined ¶ added in v0.2.0
Combined returns SystemPrompt followed by UserPrompt separated by a blank line when both are non-empty. Runners without native system-prompt support use this to flatten the Request into a single payload. The hash used for summary-skip detection is computed over this combined form so changes to either half invalidate the cached summary.
type RunResult ¶
type RunResult struct {
Text string
Meta *LLMMetadata
}
RunResult holds the LLM response text and optional metadata.
type Runner ¶
Runner executes a structured LLM request and returns the response with metadata. The implementation decides which model and transport to use. Injected so tests can substitute fakes.
type Severity ¶
type Severity string
Severity classifies a pre-flight finding. Mirrored in the query package; templates describe severity in purely semantic terms.
type SummarizeResult ¶
SummarizeResult holds the generated summary and its prompt hash.
func Summarize ¶
func Summarize(ctx context.Context, runner Runner, entry *model.Entry, graph *model.Graph, force bool) (*SummarizeResult, error)
Summarize generates a summary for a single entry using the LLM runner. Returns nil if the entry's stored SummaryHash matches the computed hash (skip). Set force to regenerate regardless.
Source Files
Directories
¶
| Path | Synopsis |
|---|---|
|
Package claude implements llm.Runner by invoking the Claude CLI.
|
Package claude implements llm.Runner by invoking the Claude CLI. |
|
Package factory resolves an llm.Runner from model.LLMConfig.
|
Package factory resolves an llm.Runner from model.LLMConfig. |
|
Package gollm implements llm.Runner on top of github.com/teilomillet/gollm, providing a unified adapter for Anthropic API, OpenAI, Ollama, and other providers supported by gollm.
|
Package gollm implements llm.Runner on top of github.com/teilomillet/gollm, providing a unified adapter for Anthropic API, OpenAI, Ollama, and other providers supported by gollm. |