fc-plugin-codbi/com.github.xima_formcycle_entwicklerkreis.fc.plugin.codbi.logic.cb.ai.llama/Standard

Standard

Standard — Generic GGUF model runner via local LLAMA-Server process. All AI computation happens in the external LLAMA-Server process. If it OOMs the Tomcat JVM stays alive — only the LLAMA-Server dies.

Plugin Properties

Property	Type	Default	Description
`Active_AI`	String	—	Must contain `llama_std` to activate this model
`AI_LLAMA_STD_ModelUrl`	URL	Qwen3-VL-2B Q4_K_M HuggingFace	Download URL for the GGUF model file
`AI_LLAMA_STD_MmprojUrl`	URL	Qwen3-VL-2B mmproj (when using default model)	Download URL for the vision projector (mmproj) file. Omit for text-only models (vision features disabled). Auto-set when using the default VL model
`AI_LLAMA_STD_MaxPixels`	Long	`3211264`	Max pixel budget for image downscaling (min 3136)
`AI_LLAMA_STD_MaxUploadBytes`	Long	`52428800`	Max raw image size in bytes before decoding (default 50 MB, min 1 MB)
`AI_LLAMA_STD_MaxTokens`	Int	`2048`	Maximum tokens to generate per response
`AI_LLAMA_STD_MaxRAMPercent`	Double	`101.0`	RAM usage threshold (%) — blocks requests when exceeded
`AI_LLAMA_STD_MaxComputePercent`	Double	`101.0`	Compute usage threshold (%) — gates on GPU% (CUDA) or CPU% (fallback). Blocks requests when exceeded
`AI_LLAMA_STD_MaxCPUPercent`	Double	—	Legacy alias for MaxComputePercent (accepted as fallback)
`AI_LLAMA_STD_LlamaRelease`	String	`b8175`	llama.cpp release tag for server binary download
`AI_LLAMA_STD_ServerUrl_<platform>`	URL	(auto from release tag)	Per-platform override for the LLAMA-Server binary URL
`AI_LLAMA_STD_UpdateCheckHours`	Long	`24`	Hours between GitHub release checks (0 = disabled)
`AI_LLAMA_STD_NotifyEmail`	String	—	Email address for update notifications
`AI_LLAMA_STD_ThinkingModelUrl`	URL	—	Download URL for a dedicated thinking model GGUF (optional)
`AI_LLAMA_STD_ThinkingMmprojUrl`	URL	—	Download URL for the thinking model's mmproj file (optional)
`AI_LLAMA_STD_ExternalUrl`	URL	—	Base URL of an external OpenAI-compatible API; overrides local model
`AI_LLAMA_STD_ExternalApiKey`	String	—	API key for the external AI (sent as Bearer token)
`AI_LLAMA_STD_ExternalModel`	String	—	Model name for the external API (e.g. gpt-4o, claude-3-opus)
`AI_LLAMA_STD_ExternalNoPrompt`	Boolean	`false`	When `true`, skips all built-in system-prompt sections (§1–§6) for the external AI — sends only the user message and chat history.
`AI_LLAMA_STD_PromptIdentity`	String	(built-in)	Override the identity/role sentence ("You are a helpful assistant..."). Use `{date}` for today's date, `{time}` for current time.
`AI_LLAMA_STD_PromptLocation`	String	(built-in)	Override the location-context instruction. Use `{location}` as placeholder.
`AI_LLAMA_STD_PromptSearch`	String	(built-in)	Override the CALL:search instruction block (before examples).
`AI_LLAMA_STD_PromptThinking`	String	(built-in)	Override the thinking-mode instruction. Use `{language}` as placeholder.
`AI_LLAMA_STD_PromptNoInternet`	String	(built-in)	Override the no-internet-access warning.
`AI_LLAMA_STD_PromptRules`	String	(built-in)	Override the general rules (language, measurements, independence).
`AI_LLAMA_STD_FallbackLocation`	String	—	Fallback location string used when geolocation fails (e.g. `Ansbach, Nürnberger Straße 32, Bayern, Deutschland`)
`AI_LLAMA_STD_NominatimDomain`	String	`nominatim.openstreetmap.org`	Domain for reverse geocoding requests (without path).
`AI_LLAMA_STD_IpGeolocationDomain`	String	`ipwho.is`	Domain for IP geolocation requests (without path).
`AI_BraveSearch_ApiKey`	String	—	Brave Search API key — enables web search tool for the model
`AI_BraveSearch_MaxResults`	Int	`5`	Maximum number of Brave Search results per query (1–20).
`AI_LLAMA_STD_Language`	String	—	Two-letter ISO 639-1 code (e.g. `de`, `fr`) — forces the AI to respond in this language, skipping auto-detection. Overridden by per-functionality `responselanguage` toLoad.
`AI_LLAMA_STD_SPECIALIST_XXX`	URL	—	Download URL for a specialist GGUF model named `XXX`. The name is chosen by the administrator and matched case-insensitively by the `specialist` toLoad property.
`AI_LLAMA_STD_SPECIALIST_MMProj_XXX`	URL	—	Download URL for the specialist `XXX`'s multimodal projector (mmproj). Optional — omit if the specialist model has no vision capability.
`AI_LLAMA_STD_EXT_SPECIALIST_XXX`	URL	—	Base URL of an external OpenAI-compatible API for a specialist named `XXX`. Matched case-insensitively by the `specialist` toLoad property.
`AI_LLAMA_STD_EXT_SPECIALIST_Key_XXX`	String	—	API key for the external specialist `XXX` (sent as Bearer token). Optional.
`AI_LLAMA_STD_EXT_SPECIALIST_Model_XXX`	String	—	Model name for the external specialist `XXX` (e.g. `gpt-4o`). Optional — omit to use the API default.
`AI_LLAMA_STD_ExtraParams`	JSON	—	Extra parameters appended to every completion request body (e.g. `{"top_p":0.9,"seed":42}`). Keys `messages`, `stream`, `model`, `id_slot`, `logprobs` are silently ignored.

Domains to whitelist

github.com — LLAMA-Server binary releases & release-check API
api.github.com — latest-release version checks
objects.githubusercontent.com — GitHub release asset CDN
huggingface.co — GGUF model & mmproj downloads
nominatim.openstreetmap.org — reverse geocoding for location context
ipwho.is — IP-based geolocation fallback
api.search.brave.com — Brave web search (only when AI_BraveSearch_ApiKey is configured)

Constructors

Standard

constructor()

Types

Companion

object Companion

Companion for static members.

Functions

execute

open override fun execute(params: IPluginServletActionParams): IPluginServletActionRetVal

Routes incoming requests to one of three handlers:

getDescription

open fun ~~getDescription~~(): String

open fun getDescription(p0: Locale): String

getDisplayName

open override fun getDisplayName(p0: Locale): String

getName

open override fun getName(): String

initialize

open override fun initialize(configData: IPluginInitializeData)

Reads all plugin properties, downloads model files if needed, launches the LLAMA-Server process, and starts the resource monitor and version-check daemon.

initPlugin

open fun ~~initPlugin~~()

install

open fun install(p0: IPluginInstallData)

shutdown

open override fun shutdown(shutdownData: IPluginShutdownData?)

Tears down all background resources: update checker, resource monitor, streaming sessions, thinking server, and the base-class server process.

open fun ~~shutdown~~()

startJanitor

fun startJanitor()

Initiates a task that removes unused images that're expired (msExpirationIDedImages) from the cache (cacheIDedImages).

uninstall

open fun uninstall(p0: IPluginUninstallData)

validateConfigurationData

open override fun validateConfigurationData(configData: IPluginValidationData): IPluginInitializeValidationResult?

Rejects tenant-level installation. CodBi must be installed as a system plugin because its AI services (Whisper, LLAMA) bind local server ports and manage heavyweight processes that would conflict when instantiated once per tenant.