HQ Radio Knowledge Base Conversation Catalog

HQ Radio Knowledge-Base Conversation Catalog (kbTell sentences, FSM reactions)

Source-verified 2026-06-21 against master 0139a346. Paths relative to Missions/[55-2hc]warfarev2_073v48co.chernarus/ unless noted (other roots: Server/). Arma 2 OA 1.64.

Warfare's HQ radio announcer speaks through Arma 2's knowledge-base conversation system, not through playSound/sideChat. A logic-side announcer object (wfbe_radio_hq) is given an identity and a knowledge-base topic via kbAddTopic; gameplay code then calls unit kbTell [receiver, topic, sentenceClass, ...args, true], and the engine renders the sentence's text = line into the radio subtitle while playing the per-word speech[] samples from the faction's RadioAnnouncers_Config. This page catalogs the three files that define that system — Client/kb/hq.bikb (the sentence database), Client/kb/hq.fsm (the reaction FSM), Client/kb/hq.sqf (the topic callback) — plus the kbAddTopic registration and every kbTell callsite.

This is the client-side spoken-radio half. The server-side payload table that decides which sentence class fires for which game event (the SideMessage switch over Lost/Captured/Constructed/...) lives in Server/Functions/Server_SideMessage.sqf and is documented as the "SideMessage Pipeline" in Server-Gameplay-Runtime-Atlas; the bottom of this page maps the two together.

The three knowledge-base files

File	Role	Size
Client/kb/hq.bikb	Sentence database — class Sentences of spoken-line classes (config text, not SQF)	173 lines
Client/kb/hq.fsm	WF_HQ reaction FSM run when a sentence is received (BIS-FSM-editor config, not SQF)	109 lines
Client/kb/hq.sqf	Topic callback compiled at registration — empty (single blank line)	1 line

The hq.sqf callback (Client/kb/hq.sqf:1) is an empty stub. It is the fourth argument to every kbAddTopic call and is compiled via call compile preprocessFileLineNumbers "Client\kb\hq.sqf", but since the file is blank the topic has no per-sentence script callback — all reaction behaviour comes from the FSM, and the meaningful per-sentence rendering is the text/speech[] from the .bikb.

Registration: kbAddTopic

The same topic id, .bikb, .fsm and (empty) .sqf callback are registered three times — twice on the client (the announcer object and the player both need the topic so the engine can speak and the player can receive), and once on the server when it builds the per-side announcer.

Site (path:line)	Registrant	Topic id
Client/Init/Init_Client.sqf:458	_HQRadio (the wfbe_radio_hq announcer logic)	WFBE_V_HQTopicSide
Client/Init/Init_Client.sqf:459	player	WFBE_V_HQTopicSide
Server/Init/Init_Server.sqf:491	_radio_hq1 (per-side announcer)	_radio_hq_id

Client setup, in order: _HQRadio = WFBE_Client_Logic getVariable "wfbe_radio_hq" (Client/Init/Init_Client.sqf:450); WFBE_V_HQTopicSide = WFBE_Client_Logic getVariable "wfbe_radio_hq_id" (:453); _HQRadio setIdentity WFBE_V_HQTopicSide (:455); setRank "COLONEL" (:456); setGroupId ["HQ"] (:457); the two kbAddTopic calls (:458-459); then sideHQ = _HQRadio (:460) — sideHQ is the variable every client kbTell callsite names as the speaker.

Server side, the announcer identity is a random pick from the faction's announcer pool: _radio_hq_id = (_announcers) select floor(random (count _announcers)) (Server/Init/Init_Server.sqf:485), then the same setIdentity/setRank "COLONEL"/setGroupId ["HQ"]/kbAddTopic sequence (:488-491), and _logik setVariable ["wfbe_radio_hq_id", _radio_hq_id, true] (:492) publishes it so the client reads the matching id at Init_Client.sqf:453.

Sentence database — Client/kb/hq.bikb

The file is a class Sentences block (hq.bikb:27) of sentence classes. Each class has a text = subtitle string (literal or a $STR... stringtable key) and a speech[] = array of per-word sample tokens resolved against the faction RadioAnnouncers_Config. %1/%2/%3/%4 are positional argument slots filled by the kbTell call; each filled slot is declared in a class Arguments sub-block of type = "simple".

Macros

Three preprocessor macros generate the bulk of the one-word and keyed sentences:

Macro (path:line)	Expands to	Subtitle source	speech[] shape
SENTENCE(NAME)	hq.bikb:1-7	text = $STR<NAME> (stringtable key STR + class name)	{%1, NAME, %2} (announcer word sits between two arg slots)
SENTENCE1(NAME)	hq.bikb:9-15	text = $STR<NAME>	{NAME, %1, %2} (announcer word first)
SENTENCE_KEY(NAME,KEY)	hq.bikb:18-24	text = $<KEY> (explicit stringtable key)	{%1, NAME, %2}

Explicit (hand-written) sentence classes

These are spelled out in full rather than macro-generated. The text column is the literal subtitle template as it appears in the file.

Class (path:line)	text subtitle	speech[] words	args
Test (hq.bikb:30)	Test	{lopotev_present_in_hideout_1_R_1}	none — leftover test class
HQ (hq.bikb:34)	$STR_DN_WARFARE_HQ_BASE_UNFOLDED	{HQ}	none
CapturedNear (hq.bikb:36)	%1 captured near %2	{%1, CapturedNear, %2}	1, 2
LostAt (hq.bikb:46)	%1 lost at %2	{%1, LostAt, %2}	1, 2
OrderSent (hq.bikb:52)	%1 %2, %3	{%1, %2, %3}	1, 2, 3, 4
OrderSentAll (hq.bikb:63)	%1, %2, %3	{%1, %2, %3}	1, 2, 3
OrderDone (hq.bikb:73)	%1, %2 %3 %4	{%1, %2, %3, %4}	1, 2, 3
simple (hq.bikb:83)	%1	{%1}	1
ExtractionTeam (hq.bikb:91)	This is HQ, %1 is on the way, out.	{ThisIs, HQ, %1, Is, OnTheWay, Out}	1
ExtractionTeamCancel (hq.bikb:99)	This is HQ, Abort the op, %1 is in danger, out.	{ThisIs, HQ, aborttheop, %1, Is, In, DangerE, Out}	1
MMissionFailed (hq.bikb:107)	This is HQ, Mission Failure, carry on with your prior tasks, over.	{ThisIs, HQ, MissionFailure, CarryOnWithYourPriorTasksOver}	none
MMissionComplete (hq.bikb:113)	This is HQ, %1	{ThisIs, HQ, %1}	1

Macro-generated sentence classes

All resolve through the macros above. SENTENCE/SENTENCE1 keys derive from the class name ($STR<name>); SENTENCE_KEY names an explicit stringtable key. Grouped by use, in file order.

Class (path:line)	Macro	Subtitle key
10MinutesLeft (hq.bikb:122)	SENTENCE	$STR10MinutesLeft
20MinutesLeft (hq.bikb:123)	SENTENCE	$STR20MinutesLeft
5MinutesLeft (hq.bikb:124)	SENTENCE	$STR5MinutesLeft
Accomplished (hq.bikb:125)	SENTENCE	$STRAccomplished
BaseUnderAtack (hq.bikb:127)	SENTENCE_KEY	$STRwfbaseunderattack
Cancelled (hq.bikb:128)	SENTENCE	$STRCancelled
CanDoWereOnIt (hq.bikb:129)	SENTENCE	$STRCanDoWereOnIt
Captured (hq.bikb:130)	SENTENCE	$STRCaptured
HC_OrderReceived (hq.bikb:131)	SENTENCE_KEY	$Order Received
CiviliansUnderEnemyFireIn (hq.bikb:132)	SENTENCE1	$STRCiviliansUnderEnemyFireIn
Constructed (hq.bikb:133)	SENTENCE_KEY	$STRwfstructureconstructed
Deployed (hq.bikb:134)	SENTENCE	$STRDeployed
Destroyed (hq.bikb:135)	SENTENCE_KEY	$STRwfbasestructuredestroyed
EnemyBaseLocated (hq.bikb:136)	SENTENCE	$STREnemyBaseLocated
EnemyForcesDetected (hq.bikb:137)	SENTENCE	$STREnemyForcesDetected
Failed (hq.bikb:138)	SENTENCE	$STRFailed
HostilesDetectedNear (hq.bikb:139)	SENTENCE1	$STRHostilesDetectedNear
Insufficient (hq.bikb:140)	SENTENCE1	$STRInsufficient
IsUnderAttack (hq.bikb:141)	SENTENCE_KEY	$STRwftownunderattack
Lost (hq.bikb:143)	SENTENCE	$STRLost
Mobilized (hq.bikb:145)	SENTENCE	$STRMobilized
NegativeWeCannotDivert (hq.bikb:146)	SENTENCE	$STRNegativeWeCannotDivert
NewIntelAvailable (hq.bikb:147)	SENTENCE	$STRNewIntelAvailable
NewMissionAvailable (hq.bikb:148)	SENTENCE	$STRNewMissionAvailable
NewSupportAvailable (hq.bikb:149)	SENTENCE	$STRNewSupportAvailable
UnderAttack (hq.bikb:151)	SENTENCE_KEY	$STRisunderattack
VotingForNewCommander (hq.bikb:152)	SENTENCE_KEY	$STRwfvoteforcommanderinprogress

The trailing block (hq.bikb:154-168) is the structure/resource noun vocabulary used as the %1 fill-word ([_value]) of Constructed/Destroyed/etc. — not standalone announcements:

Class (path:line)	Macro	Subtitle key
AntiAirRadar (hq.bikb:154)	SENTENCE	$STRAntiAirRadar
ArtilleryBattery (hq.bikb:155)	SENTENCE	$STRArtilleryBattery
ArtilleryRadar (hq.bikb:156)	SENTENCE	$STRArtilleryRadar
Barracks (hq.bikb:157)	SENTENCE_KEY	$STRwfbarracks
Funds (hq.bikb:158)	SENTENCE	$STRFunds
Headquarters (hq.bikb:159)	SENTENCE	$STRHeadquarters
HeavyVehicleSupply (hq.bikb:160)	SENTENCE	$STRHeavyVehicleSupply
Helipad (hq.bikb:161)	SENTENCE	$STRHelipad
LightVehicleSupply (hq.bikb:162)	SENTENCE	$STRLightVehicleSupply
Mission (hq.bikb:163)	SENTENCE	$STRMission
Resources (hq.bikb:164)	SENTENCE	$STRResources
ServicePoint (hq.bikb:165)	SENTENCE	$STRServicePoint
Strongpoint (hq.bikb:166)	SENTENCE	$STRStrongpoint
Supplies (hq.bikb:167)	SENTENCE	$STRSupplies
UAVTerminal (hq.bikb:168)	SENTENCE	$STRUAVTerminal

Trailer (outside class Sentences)

Declaration (path:line)	Meaning
class Arguments{} (hq.bikb:170)	Empty top-level args block — conversation-system boilerplate
class Special{} (hq.bikb:171)	Empty Special topic class — boilerplate; not referenced by any kbTell callsite in Client/Server/Common
startWithVocal[] = {hour} (hq.bikb:172)	Article-elision hint (a/an)
startWithConsonant[] = {europe, university} (hq.bikb:173)	Article-elision hint

Reaction FSM — Client/kb/hq.fsm

fsmName = "WF_HQ" (hq.fsm:24), compiled from the BIS FSM editor (header WF_HQ, hq.fsm:1). It runs on the receiver when a sentence arrives, with _sentenceId/_from/_topic in scope. The graph is a near-verbatim copy of the campaign CooperGalkinaScene reaction template — its conditions key on sentence ids (Cooper_ComeOnWithMe, Interrupted) that no Warfare kbTell callsite ever sends, so in normal play the React state immediately falls through its true-priority link to END. It is effectively an inert/no-op reaction graph kept as a stub. initState = "React" (hq.fsm:102); finalStates[] = {"END", "___Nervous_anim", "Come_on__come"} (hq.fsm:103-108).

State (path:line)	Role	Links (to / condition)
React (hq.fsm:28)	Entry. init debug-logs the sentence if BIS_DEBUG_DIALOG is set (hq.fsm:31); otherwise inert	→ ___Nervous_anim if _sentenceId in ["Cooper_ComeOnWithMe"] (prio 1, :39-41); → Come_on__come if _sentenceId in ["Interrupted"] (prio 1, :49-51); → END on true (prio 0, :59-61)
END (hq.fsm:69)	Terminal no-op (init = "")	none
___Nervous_anim (hq.fsm:80)	Galkina nervous-escape scene (spawns animation, rKBTELL Galkina_NoNoMore, hq.fsm:83) — dead campaign code, never reached in Warfare	none
Come_on__come (hq.fsm:91)	Sends rKBTELL ... Cooper_ComeOnWithMe (hq.fsm:94) — dead campaign code, never reached	none

Because every React transition except the unconditional → END requires a campaign-only sentence id, the practical behaviour is: receive sentence → (optional debug log) → END. The HQ radio's user-visible output is entirely the .bikb subtitle + speech samples; the FSM adds nothing in Warfare.

kbTell trigger sites

Two distinct producers fire sentences from this topic: player-initiated order chatter (client GUI / task code, speaker = sideHQ, receiver = player) and the server SideMessage dispatcher (speaker/receiver = the per-side radio logic). Every call ends with the trailing true (the "force radio" / commit flag).

Player-initiated (client)

Site (path:line)	Sentence class	Trigger
Client/GUI/GUI_Menu_Command.sqf:296	OrderSent	Move order to a specific team set (Team ..., MenuAction move branch, !_isAll)
Client/GUI/GUI_Menu_Command.sqf:303	OrderSentAll	Move order to all teams (_isAll)
Client/GUI/GUI_Menu_Command.sqf:330	OrderSent	Set-Task order (MenuAction 306) to a specific team set
Client/GUI/GUI_Menu_Command.sqf:340	OrderSentAll	Set-Task order (MenuAction 306) to all teams
Client/PVFunctions/SetTask.sqf:34	OrderDone	A SimpleTask completes (_succeed) — squad reports "HQ, We are ready for orders, over."

Example argument shape (GUI_Menu_Command.sqf:296): player kbTell [sideHQ, WFBE_V_HQTopicSide, "OrderSent", ["1","",_sideTeam,[if (sideJoined == west) then {"blueTeam"} else {"redTeam"}]], ["2","","moving to position",["HC_MovingToPosition"]], ["3","","over.",["Over1"]], true] — each ["idx","",displayString,[speechToken]] tuple fills one positional %n with a subtitle fragment and its spoken word.

Server SideMessage dispatcher

Server/Functions/Server_SideMessage.sqf is the bound SideMessage function (SideMessage = Compile preprocessFile "Server\Functions\Server_SideMessage.sqf", Server/Init/Init_Server.sqf:39). It resolves the per-side radio logic (wfbe_radio_hq speaker, wfbe_radio_hq_rec receiver, wfbe_radio_hq_id topic, :8-12), guards against a base-less faction (if (isNull _speaker) exitWith {}, :10), then switch (true) selects the kbTell form by message class:

Site (path:line)	Sentence classes handled	Argument fill
Server/Functions/Server_SideMessage.sqf:22	Lost, Captured, HostilesDetectedNear	one arg: town/object real name (_rlName) with _dub speech token
:31	CapturedNear, LostAt	two args: capturing-side label + the town name/dub
:57	Constructed, Destroyed, Deployed, Mobilized, IsUnderAttack	one arg: localized structure noun + its vocabulary token (Headquarters/Barracks/LightVehicleSupply/...)
:60	VotingForNewCommander, NewIntelAvailable, MMissionFailed, NewMissionAvailable	no args (bare kbTell [_receiver,_topicSide,_message,true])
:63	MMissionComplete, ExtractionTeam, ExtractionTeamCancel	one arg: _parameters select 0 text with select 1 speech token

The Constructed/Destroyed branch (:34-56) is where the structure-noun vocabulary block from the .bikb is consumed: it maps wfbe_structure_type (Headquarters/Barracks/Light/CommandCenter/Heavy/Aircraft/ServicePoint/AARadar) to a localized string + a _value token that becomes the %1 speech word. Note CommandCenter resolves its spoken token to UAVTerminal, switching to CommandPost for Arrowhead / west-CombinedOps (:43-45).

How the two halves connect

                 client GUI / task                  server game events
                 (move/task orders)                 (town capture, build, votes...)
                       |                                     |
   player kbTell [sideHQ, topic, "OrderSent"...]   [side,"Lost",town] call SideMessage
                       |                                     |
                       |                            Server_SideMessage.sqf switch
                       |                              picks sentence class + args
                       +------------------+------------------+
                                          |
                          hq.bikb  class <Name> { text=...; speech[]=... }
                                          |
                          engine renders subtitle + plays speech[] words
                                          |
                          receiver runs hq.fsm (WF_HQ) -> React -> END (no-op)
                                          |
                          hq.sqf topic callback (empty) -> nothing

The .bikb is the shared vocabulary both producers draw from; the client GUI path and the server SideMessage path are independent triggers into it. The FSM and .sqf callback are stubs that add no behaviour in Warfare — the radio's entire observable output is the .bikb text/speech[] rendering. For the event→class decision table on the server side (which game event maps to Lost vs CapturedNear vs Constructed), see the SideMessage section of Server-Gameplay-Runtime-Atlas.

Muting the command voice (per-player toggle — design note)

Because the entire HQ/command voice is the engine's radio-protocol sentence system (kbTell) and nothing else — the only three kbTell producers are Server/Functions/Server_SideMessage.sqf, Client/GUI/GUI_Menu_Command.sqf, and Client/PVFunctions/SetTask.sqf — a per-player on/off toggle is straightforward and collateral-free.

• Engine mechanism: Arma 2 OA's enableSentences false disables the radio-protocol sentence channel that kbTell speaks through (enableSentences true re-enables it). It is a local-effect command, so a player who calls it mutes the command voice only on their own machine, affecting no one else. enableRadio false is the analogous toggle for formatted radio messages; the HQ voice is sentences, so enableSentences is the precise lever.
• No existing toggle / no conflict: neither enableSentences nor enableRadio is called anywhere in the mission (grep over *.sqf → 0 hits), so the channel is fully on by default and there is currently no player-facing off switch to collide with.
• Scope caveat: enableSentences false suppresses the whole sentence (spoken samples and its subtitle) — it is a true mute, not a voice-only/keep-subtitles option. Keeping subtitles while muting audio has no engine toggle and would require gating delivery (and Server_SideMessage broadcasts side-wide), so it is materially harder.
• Drop-in implementation pattern: mirror the existing Earplugs client QoL toggle — WASP/actions/EarplugToggle.sqf (which already dampens radio to 12 % via 0.25 fadeRadio 0.12 at EarplugToggle.sqf:17, a combined ambient+radio earplug, not a dedicated voice mute) registered and respawn-re-applied through WASP/actions/AddActions.sqf:33 onward. A dedicated voice toggle would be a sibling WASP scroll-menu action that flips a missionNamespace flag and calls enableSentences, with the preference persisted across sessions in profileNamespace and re-applied at client init so it survives JIP/respawn.

This is a design note, not shipped behaviour — no enableSentences toggle exists on master 0139a346.

Continue Reading

• Server-Gameplay-Runtime-Atlas — the server-side SideMessage payload table: which game event maps to which sentence class, the sibling of this client-side spoken catalog.
• LocalizeMessage-Chat-Notification-Router-Reference — the text-chat notification router (a separate, non-radio channel for player-facing messages).
• Mission-Audio-Catalog — the CfgSounds/CfgMusic registry and playSound cues, the other audio path distinct from kbTell speech samples.
• Commander-HQ-Lifecycle-Atlas — the HQ/commander lifecycle behind the announcer logic and the wfbe_radio_hq per-side object.
• Quad-AI-Commander — the AI commander that drives many of the server events feeding the SideMessage radio announcements.
• Earplugs-Audio-Toggle-Reference — the existing client audio QoL toggle whose action/registration pattern a per-player command-voice mute would mirror.