buildMessageText combines chat messages into a single prompt for Copilot.

captureInput captures the input for a function call

checkImageGenerationCompatibility warns if the model doesn't support image generation

convertMessageCommon extracts common conversion logic

countPatternsInDirectory counts the number of pattern directories in a given directory

createConversation creates a new Copilot conversation.

Create a new pattern with necessary files and structure.

detectSystemLocale detects the system locale using standard Unix environment variables. Follows the POSIX priority order for locale environment variab

extractModelName extracts the model name from the full resource path Input: "publishers/google/models/gemini-2.0-flash" Output: "gemini-2.0-flash"

extractResponseText extracts the assistant's response from messages.

Fetch models once and cache the results.

generateWAVFile creates WAV data from PCM data with proper headers

Gemini methods using genai SDK with Vertex AI backend getGeminiRegion returns the appropriate region for a Gemini model. Preview models are often only

getLocaleCandidates returns a list of locale candidates to try, in order of preference. For example, for "pt-PT" it returns ["pt-PT", "pt", "pt-BR"] (

getOutputFormatFromExtension determines the API output format based on file extension

getSortedGroupsItems returns a new slice of GroupItems with both groups and their items sorted alphabetically in a case-insensitive manner. The origin

Get the directory for storing outputs.

handleSetupAndServerCommands handles setup and server-related commands Returns (handled, error) where handled indicates if a command was processed and

isConversationalModel returns true if the model is suitable for text generation/chat Filters out image generation, embeddings, and other non-conversat

isMergeCommit determines if a commit is a merge commit based on its parents and message patterns.

isTTSModel checks if the model is a text-to-speech model

isValidLocale checks if a locale string can be parsed as a valid language tag.

joinPromptSections trims each part, drops empty ones, and joins the rest with newline separators.

maskSecret redacts a secret value for display, showing only the first 4 and last 4 chars.

modelPriority returns the sort priority for a model (lower = higher priority)

newOllamaEngine makes the engine but does not start it, which lets tests operate the routes. The address parameter is the /api/chat forward target, no

newTraversalFixture returns a storage entity in a temporary root and a marker file out of the entity directory. It also returns a check that fails the

normalizeLocale converts various locale formats to BCP 47 language tags. Examples: - "en_US.UTF-8" -> "en-US" - "fr_FR@euro" -> "fr-FR" - "zh_CN.GB231

normalizeToBCP47 normalizes a locale string to BCP 47 format. Converts underscores to hyphens and ensures proper casing (language-REGION).

parseGeminiThinking converts thinking level to Gemini thinking config

parseOllamaNumCtx extracts and validates the num_ctx parameter from Ollama request options. Returns: - (0, nil) if num_ctx is not present or is null -

parseTimestampToSeconds converts timestamp string (HH:MM:SS or MM:SS) to total seconds

rejectInvalidStorageName answers a 400 when name does not obey storage-name validation. An empty name passes, because the fields that this guards are

Sanitize input content by removing or replacing problematic characters. Args: input_text: The input text to sanitize Returns:

Save pattern execution log.

shouldIncludeRepeat determines if repeated content should be included based on time gap

shouldUseImageGeneration determines if image generation should be enabled This is a heuristic based on the presence of --image-file flag

sortModels sorts models by priority: Gemini > Claude > Others Within each group, models are sorted alphabetically

supportsResponsesAPI determines if the provider supports the new Responses API

symlinkContained rejects an entry at path if the entry resolves, through symlinks, to a target outside absDir. A missing entry passes, because the lex

toMessages converts the array of input messages from the ChatCompletionMessageType to the Bedrock Converse Message type. The system role messages are

tryMethodYtDlpInternal is a helper function to reduce duplication between tryMethodYtDlp and tryMethodYtDlpWithTimestamps.

validateImageFile validates the image file path and extension

validateImageParameters validates image generation parameters

Validate input content for potentially problematic characters or patterns. Args: input_text: The input text to validate Returns:

Validate a pattern's structure and content.

writeOllamaResponseStruct marshals the provided OllamaResponse and writes it as newline-delimited JSON to the HTTP response stream.

Apply executes file operations: - read:PATH - Read entire file content - tail:PATH|N - Read last N lines - exists:PATH - Check if file exists - size:P

Apply executes the requested text operation on the provided value

Apply executes fetch operations: - get:URL: Fetches content from URL, returns text content

Apply executes datetime operations with the following formats: Time: now (RFC3339), time (HH:MM:SS), unix (timestamp) Hour: startofhour, endofhour Dat

AuthHeader returns the Bedrock auth header (Bearer token)

AuthHeader returns the header name and value for authentication. Each APIM backend uses a different auth header: Bedrock: "Authorization", "Beare

AuthHeader returns the Vertex AI auth header (Google API key via APIM)