@zds-ai/cli 0.1.7 → 0.1.10

package/dist/bin/generate_image_sd.sh

@@ -2,10 +2,11 @@
 # generate_image_sd.sh - Generate image via SD API, save base64/JSON/PNG
 
 # Find config from ZDS_AI_AGENT_CONFIG_FILE
-[[ ! -s "$ZDS_AI_AGENT_CONFIG_FILE" ]] && echo Failed to find config file >&2 # && exit 1 # uncommented for testing
-LOGDIR=${ZDS_AI_AGENT_LOGS_DIR:-~/Documents/ZDS-Agents/logs}
-CURRENT_DIR="$(pwd -P)"
-LOGFILE=${LOGDIR}/${ZDS_AI_AGENT_SESSION:-no-session}.log.txt
+[[ ! -s "$ZDS_AI_AGENT_CONFIG_FILE" ]] && echo Failed to find config file $ZDS_AI_AGENT_CONFIG_FILE >&2 && exit 1
+[[ ! -s "$ZDS_AI_AGENT_LOG_FILE" ]] && echo Failed to find log file $ZDS_AI_AGENT_LOG_FILE >&2 && exit 1
+[[ -z "$ZDS_AI_AGENT_SESSION" ]] && echo Failed to validate agent session >&2 && exit 1
+[[ -z "$ZDS_AI_AGENT_HOME_DIR" ]] && echo Failed to validate agent home dir >&2 && exit 1
+LOGFILE=${ZDS_AI_AGENT_LOG_FILE}
 
 # Load environment variables
 [[ -f ~/.env ]] && source ~/.env
@@ -45,6 +46,7 @@ if [[ "$1" == --help || "$1" == -h || $# -eq 0 ]]; then
   echo "  --seed <num>             Seed for reproducible generation (default: random)"
   echo "  --name <name>            Name to call the file (default: based on prompt)"
   echo "  --list-models            Show checkpoint models installed"
+   echo "  --get-lora-details <name> Show complete JSON details for specified LoRA"
   echo "  --help                   Show this help message"
   echo
   echo Environment Variables:
@@ -57,6 +59,7 @@ if [[ "$1" == --help || "$1" == -h || $# -eq 0 ]]; then
   echo "  ${PROG_NAME} 'sunset' --move --cfg-scale 7.5"
   echo "  ${PROG_NAME} 'whistler\'s mother' --steps 50 --cfg-scale 8.0 --name 'whistlers mom'"
   echo "  ${PROG_NAME} 'portrait' --seed 12345 --name 'reproducible-portrait'"
+   echo "  ${PROG_NAME} --get-lora-details 'RealisticSkinv2_ponyv6_loraplus'"
   exit 0
 fi
 
@@ -126,11 +129,24 @@ while [[ $# -gt 0 ]]; do
       SEED="$2"
       shift 2 # past argument and value
       ;;
+    --list-loras)
+      curl -s "$ZDS_AI_IMAGE_ENDPOINT"/loras | jq -r '.[].name'
+      exit 0
+      ;;
     --list-models)
       curl -s "$ZDS_AI_IMAGE_ENDPOINT"/sd-models | jq -r '.[].title' | \
         sed -e 's,\.safetensors.*,,gi'
       exit 0
       ;;
+    --get-lora-details)
+      LORA_NAME="$2"
+      if [[ -z "$LORA_NAME" ]]; then
+        echo "Error: Lora name is required for --get-lora-details" >&2
+        exit 1
+      fi
+      curl -s "$ZDS_AI_IMAGE_ENDPOINT"/loras | jq --arg name "$LORA_NAME" '.[] | select(.name == $name)'
+      exit 0
+      ;;
     -*|--*)
       echo "Unknown option $1"
       exit 1
@@ -150,13 +166,8 @@ PROMPT="$1"
 NEGATIVE_PROMPT="${2:-score_6, score_5, score_4, (worst quality:1.2), (low quality:1.2), (normal quality:1.2), lowres, bad anatomy, bad hands, signature, watermarks, ugly, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, extra limb, missing limbs}"
 
 # Override with command line flags if specified
-if [[ -n "$CFG_SCALE_OVERRIDE" ]]; then
-  CFG_SCALE="$CFG_SCALE_OVERRIDE"
-fi
-
-if [[ -n "$STEPS_OVERRIDE" ]]; then
-  STEPS="$STEPS_OVERRIDE"
-fi
+[[ -n "$CFG_SCALE_OVERRIDE" ]] && CFG_SCALE="$CFG_SCALE_OVERRIDE"
+[[ -n "$STEPS_OVERRIDE" ]] && STEPS="$STEPS_OVERRIDE"
 
 # Validate that we have at least a prompt
 if [[ -z "$PROMPT" ]]; then
@@ -185,7 +196,7 @@ NAME=${NAME:-${PROMPT}}
 SLUG=$(echo -n $NAME | tr '[:space:]' '_' | sed -e 's/[^a-zA-Z0-9_]/_/g' | cut -c 1-32)
 
 # Dirs
-OUTDIR=${ZDS_AI_AGENT_HOME_DIR:-~/Documents/ZDS-Agents}/out
+OUTDIR=${ZDS_AI_AGENT_HOME_DIR}/out
 mkdir -p $OUTDIR/photos/tmp
 OUTPUT_B64=$OUTDIR/photos/tmp/${SLUG}.b64
 OUTPUT_PNG=$OUTDIR/photos/${SLUG}.png

package/dist/bin/joycaption.sh

@@ -0,0 +1,37 @@
+#!/usr/bin/env zsh
+
+set -e
+
+# Check for help option
+if [[ "$1" == "--help" ]] || [[ "$1" == "-h" ]]; then
+    echo "Usage: joycaption <image_file> [--prompt <prompt_text>]"
+    echo
+    echo "Generate captions for images using the JoyCaption service."
+    echo
+    echo "Arguments:"
+    echo "  <image_file>     Path to the image file to caption"
+    echo "  --prompt         Optional prompt to guide the captioning"
+    echo
+    echo "Examples:"
+    echo "  joycaption image.jpg"
+    echo "  joycaption image.png --prompt \"Describe this image in detail\""
+    echo
+    exit 0
+fi
+
+IMAGE="$1"
+IMAGE_BASENAME=$(basename "$1" | tr -cs '[:alnum:].' _ | sed -e 's,_+$,,g')
+shift
+
+[[ "$2" == --prompt ]] && shift
+[[ -n "$2" ]] && IMAGE_PROMPT="--prompt \"$2\""
+
+# send the file to the server
+scp "$IMAGE" asrock:scm/joycaption/images/$IMAGE_BASENAME
+tmp=$(mktemp /tmp/joycaption.XXXXXX)
+ssh asrock "bin/joycaption-receiver $IMAGE_BASENAME $IMAGE_PROMPT" > $tmp 2>/dev/null
+grep -A99 '==========' $tmp | grep -v '==========' > ${tmp}.txt
+[[ -s ${tmp}.txt ]] && cat ${tmp}.txt || cat ${tmp}
+rm -f $tmp
+
+exit 0

package/dist/bin/talking-agents.sh

@@ -0,0 +1,291 @@
+#!/usr/bin/env zsh
+
+# set -x
+
+# Load environment variables
+[[ -f ~/.env ]] && source ~/.env
+
+# Find config from ZDS_AI_AGENT_CONFIG_FILE
+[[ ! -s "$ZDS_AI_AGENT_CONFIG_FILE" ]] && echo Failed to find config file $ZDS_AI_AGENT_CONFIG_FILE >&2 && exit 1
+[[ ! -s "$ZDS_AI_AGENT_LOG_FILE" ]] && echo Failed to find log file $ZDS_AI_AGENT_LOG_FILE >&2 && exit 1
+[[ -z "$ZDS_AI_AGENT_SESSION" ]] && echo Failed to validate agent session >&2 && exit 1
+LOGFILE=${ZDS_AI_AGENT_LOG_FILE}
+
+# Log the tool invocation
+{
+  echo "=== $(basename "$0") Tool Fired: $(date) ==="
+  echo Agent bot name: ${ZDS_AI_AGENT_BOT_NAME}
+  echo
+  set | grep ^ZDS_AI_AGENT_ | grep -vE ^ZDS_AI_AGENT_'(BOT_NAME=)'
+  echo
+  echo Params: "$@"
+  echo
+} >> $LOGFILE
+
+_build_voice_strings() {
+  local yq_path=$1
+  local voice_parts=()
+  local voice_remote_parts=()
+  local voice_count=$(yq "${yq_path}.kokoro.voices | length" "$ZDS_AI_AGENT_CONFIG_FILE")
+
+  for ((i=0; i<voice_count; i++)); do
+    local name=$(yq "${yq_path}.kokoro.voices[$i].name" "$ZDS_AI_AGENT_CONFIG_FILE")
+    local weight=$(yq "${yq_path}.kokoro.voices[$i].weight" "$ZDS_AI_AGENT_CONFIG_FILE")
+    voice_parts+=("${name}:${weight}")
+    voice_remote_parts+=("${name}(${weight})")
+  done
+
+  echo "$(IFS=',' ; echo "${voice_parts[*]}")|$(IFS='+' ; echo "${voice_remote_parts[*]}")"
+}
+
+_load_mode_config() {
+  local mode=$1
+
+  SPEED=$(yq ".speech.$mode.kokoro.speed" "$ZDS_AI_AGENT_CONFIG_FILE")
+  NORM=$(yq ".speech.$mode.kokoro.norm" "$ZDS_AI_AGENT_CONFIG_FILE")
+  PITCH=$(yq ".speech.$mode.kokoro.pitch" "$ZDS_AI_AGENT_CONFIG_FILE")
+
+  local voices=$(_build_voice_strings ".speech.$mode")
+  VOICE="${voices%|*}"
+  VOICE_REMOTE="${voices#*|}"
+}
+
+# Load default TTS configuration
+_load_mode_config default
+CURRENT_MODE=default
+
+SUFFIX=
+MOVE_FLAG=
+EXT=mp3
+
+# for var in SPEED NORM PITCH VOICE_REMOTE SUFFIX; do
+#   echo $var = ${(P)var} >> $LOGFILE
+# done
+
+_tts_fix_audio() {
+  ffmpeg -v error -i "$1" -af deesser,haas=side_gain=0.3,\
+arnndn=/usr/local/share/models/cb.rnnn,\
+afftdn=nr=10:nf=-80:tn=1,adynamicsmooth=2,anlmdn=m=99,\
+adynamicsmooth=99:999 -y "$2" | tee -a "$LOGFILE"
+}
+
+_tts_postprocess_audio() {
+  sox "$1" "$2" pad 0 1 norm ${NORM} pitch ${PITCH} | tee -a "$LOGFILE"
+}
+
+_tts_render_kokoro() {
+#   kokoro -s 0.77 -m af_bella -i "$1" -o "$2" || :
+#   cd ~/scm/kokoro-tts
+  kokoro-tts "$1" "$2" --speed $SPEED --voice ${VOICE} \
+    --model ~/.cache/kokoro-tts/kokoro-v1.0.onnx \
+    --voices ~/.cache/kokoro-tts/voices-v1.0.bin
+#   ERR=$?
+#   cd -
+#   return $ERR
+}
+
+_tts_get_language() {
+  echo $1 | lingua-cli -l en,es,pt | cut -c -2 | \
+    sed -e 's,^en$,en-us,gi' -e 's,^pt$,pt-br,gi'
+}
+
+# Get voice and speed for a specific language, trying language-specific mode variant first
+_get_language_voice_config() {
+  local mode=$1
+  local language=$2
+  local lang_var=$(echo "$language" | sed 's/-/_/g')
+  local lang_mode="${mode}_${lang_var}"
+
+  # Try language-specific variant first (e.g., loving_es for Spanish)
+  if [[ $(yq ".speech.$lang_mode" "$ZDS_AI_AGENT_CONFIG_FILE") != null ]]; then
+    local voices=$(_build_voice_strings ".speech.$lang_mode")
+    local speed=$(yq ".speech.$lang_mode.kokoro.speed" "$ZDS_AI_AGENT_CONFIG_FILE")
+    echo "${voices%|*}|$speed"
+  else
+    # Fall back to base mode
+    echo "$VOICE_REMOTE|$SPEED"
+  fi
+}
+
+# New function to accept content directly instead of reading from file
+_tts_render_kokoro_remote_content() {
+  local content="$1"
+  local output_file="$2"
+  local language=$(_tts_get_language "$1")
+
+  # Get language-appropriate voice and speed for current mode
+  local config=$(_get_language_voice_config "$CURRENT_MODE" "$language")
+  local use_voice="${config%|*}"
+  local use_speed="${config#*|}"
+
+  curl -s -X POST "${ZDS_AI_TTS_ENDPOINT}" \
+    -u "${ZDS_AI_TTS_USER}:${ZDS_AI_TTS_PASS}" \
+    -H "Content-Type: application/json" \
+    -H "X-Raw-Response: true" \
+    -d '{
+      "input": "'"${content}"'",
+      "model": "kokoro",
+      "voice": "'${use_voice}'",
+      "response_format": "wav",
+      "speed": '${use_speed}',
+      "stream": false,
+      "lang_code": "'${language}'",
+      "return_download_link": true
+    }' > "$output_file" 2>> "$LOGFILE"
+}
+
+# en-us, es, pt-br, hi, it, fr-fr, zh
+
+# Function to split text into sentences using awk
+_tts_split_text_into_sentences() {
+
+  # Read file and split into sentences (basic sentence splitting)
+  # This splits on periods, exclamation marks, question marks, and ellipsis
+  awk '{
+    # Split on sentence endings while keeping the punctuation
+    gsub(/[\.!?…]/, "&\n")
+    print
+  }' | \
+  awk 'NF > 0 {
+    # Clean up whitespace
+    gsub(/^[[:space:]]+|[[:space:]]+$/, "")
+    if (length($0) > 0) {
+      print $0
+    }
+  }'
+}
+
+# New function to split text into sentences and process them individually
+_tts_render_kokoro_remote_sentences() {
+  local input_file="$1"
+  local output_file="$2"
+  local temp_dir=$(mktemp -d "${2}.temp.XXXXXX") || return 1
+  local pad_amount=$(echo "scale=2; 0.7 / $SPEED" | bc)
+
+  # Split text into sentences first
+  _tts_split_text_into_sentences < "$input_file" > "$temp_dir/sentences.txt"
+
+  # Process each sentence
+  local sentence_num=0
+  local all_wav_files=()
+
+  while IFS= read -r sentence; do
+    if [[ -n "$sentence" ]]; then
+      sentence_num=$((sentence_num + 1))
+      wav_file="$temp_dir/sentence_${sentence_num}.wav"
+
+      # Process this sentence directly without writing to file
+      if _tts_render_kokoro_remote_content "$sentence" "$wav_file"; then
+        if [[ ${sentence_num} -eq 1 ]]; then
+          all_wav_files+=($wav_file)
+        else
+          sox $wav_file ${wav_file}.padded.wav pad ${pad_amount} 0
+          all_wav_files+=($wav_file.padded.wav)
+        fi
+      fi
+    fi
+  done < "$temp_dir/sentences.txt"
+
+  if [[ ${sentence_num} -eq 0 ]]; then
+    rm -rf $temp_dir
+    return 1
+  fi
+
+  sox $all_wav_files "$output_file" rate 48k
+  rm -rf $temp_dir
+  return 0
+}
+
+_tts_prep_file() {
+  FILE_TYPE=$(file -b "$1")
+  echo $FILE_TYPE
+  echo $FILE_TYPE | grep -q -s CRLF.\*line.terminators && dos2unix "$1"
+  if echo $FILE_TYPE | grep -q -s Non-ISO.extended-ASCII; then
+    iconv -c -f macroman -t ascii < "$1" > "$1.ascii"
+    cat "$1.ascii" > "$1"
+    rm "$1.ascii"
+  fi
+  # Sanitize text for TTS using pandoc (auto-detect format)
+  pandoc -t plain --wrap=none "$1" | \
+    tr -cd '[:alnum:][:space:][:punct:]\n' > "$1.tmp" && mv "$1.tmp" "$1"
+}
+
+_die() {
+  echo $* >&2
+  exit 1
+}
+
+if [[ "$1" == --move ]]; then
+  MOVE_FLAG=true
+  shift
+fi
+
+# Check if a custom mode is defined in config
+if [[ -n "$1" && "$1" != --move && "$1" != slow && "$1" != fast && ! -f "$1" ]]; then
+  if [[ $(yq ".speech.$1" "$ZDS_AI_AGENT_CONFIG_FILE") != null ]]; then
+    _load_mode_config "$1"
+    CURRENT_MODE="$1"
+    SUFFIX=-$1
+    shift
+  fi
+fi
+
+if [[ "$1" == slow ]]; then
+  SPEED=$(echo "$SPEED * 0.9" | bc -S 2 | awk '{printf "%f", $0}')
+  NORM=$((NORM - 2))
+  PITCH=$((PITCH - 40))
+  SUFFIX="${SUFFIX}-slow"
+  shift
+fi
+
+if [[ "$1" == fast ]]; then
+  SPEED=$(echo "$SPEED * 1.2" | bc -S 2 | awk '{printf "%f", $0}')
+  NORM=$((NORM + 2))
+  PITCH=$((PITCH + 40))
+  SUFFIX="${SUFFIX}-fast"
+  shift
+fi
+
+local outdir=${ZDS_AI_AGENT_HOME_DIR:+$ZDS_AI_AGENT_HOME_DIR/}out/Speech/
+mkdir -p ${outdir} || _die could not create ${outdir}
+[[ -d "$ZDS_AI_TTS_MOVE_DIR" ]] && MOVE_DIR="$ZDS_AI_TTS_MOVE_DIR" || \
+  MOVE_DIR="$outdir"
+
+# for var in SPEED NORM PITCH VOICE_REMOTE SUFFIX MOVE_DIR; do
+#   echo $var = ${(P)var} >> $LOGFILE
+# done
+
+# for a in pre/*(N); do
+for a in "$@"; do
+  [[ "$a" == --move ]] && MOVE_FLAG=true && continue
+
+  b=${outdir}$(basename "$a" .txt)$SUFFIX.$$
+  (
+    _tts_prep_file "$a"
+#    _tts_render_kokoro "$a" "${b}-temp1.wav" || continue
+    _tts_render_kokoro_remote_sentences "$a" "${b}-temp1.wav" || continue
+    _tts_fix_audio "${b}-temp1.wav" "${b}-temp2.wav" || continue
+    _tts_postprocess_audio "${b}-temp2.wav" "${b}.wav" || continue
+    rm -f "${b}-temp1.wav" "${b}-temp2.wav"
+    echo "${b}.wav"
+    soxi "${b}.wav"
+    sox "${b}.wav" "${b}.$EXT"
+  ) > ${b}.stdout.txt 2> ${b}.stderr.txt
+  if [[ -s "${b}.$EXT" ]]; then
+    rm "${b}.stdout.txt"
+    rm "${b}.stderr.txt"
+    if [[ $MOVE_FLAG == true ]]; then
+      mv -v "${b}.$EXT" "$MOVE_DIR"
+      rm "${b}.wav"
+      echo "$MOVE_DIR"/$(basename "${b}.$EXT")
+    else
+      echo "${b}.$EXT"
+    fi
+  else
+    cat "${b}.stdout.txt"
+    cat "${b}.stderr.txt" >&2
+    exit 1
+  fi
+done
+
+exit 0

package/dist/grok/client.d.ts

@@ -20,11 +20,42 @@ export interface LLMToolCall {
         arguments: string;
     };
 }
+/**
+ * @deprecated Use WebSearchTool with the Agent Tools API instead.
+ * The search_parameters field is deprecated as of December 15, 2025.
+ */
 export interface SearchParameters {
     mode?: "auto" | "on" | "off";
 }
+export interface WebSearchTool {
+    type: "web_search";
+    allowed_domains?: string[];
+    excluded_domains?: string[];
+    enable_image_understanding?: boolean;
+    user_location?: {
+        type: "approximate";
+        country?: string;
+        city?: string;
+        region?: string;
+        timezone?: string;
+    };
+}
 export interface SearchOptions {
+    /** @deprecated Use enable_web_search and web_search_filters instead */
     search_parameters?: SearchParameters;
+    enable_web_search?: boolean;
+    web_search_filters?: {
+        allowed_domains?: string[];
+        excluded_domains?: string[];
+        enable_image_understanding?: boolean;
+        user_location?: {
+            type: "approximate";
+            country?: string;
+            city?: string;
+            region?: string;
+            timezone?: string;
+        };
+    };
 }
 export interface LLMResponse {
     choices: Array<{
@@ -43,8 +74,10 @@ export declare class LLMClient {
     private backendName;
     private supportsTools;
     private supportsVision;
+    private disableReasoningOverride;
     constructor(apiKey: string, model?: string, baseURL?: string, displayName?: string);
     setModel(model: string): Promise<void>;
+    private parseModelSuffix;
     private enableTools;
     private enableVision;
     getCurrentModel(): string;
@@ -55,6 +88,27 @@ export declare class LLMClient {
     setSupportsVision(value: boolean): void;
     chat(messages: LLMMessage[], tools?: LLMTool[], model?: string, searchOptions?: SearchOptions, temperature?: number, signal?: AbortSignal, maxTokens?: number): Promise<LLMResponse>;
     chatStream(messages: LLMMessage[], tools?: LLMTool[], model?: string, searchOptions?: SearchOptions, temperature?: number, signal?: AbortSignal, maxTokens?: number): AsyncGenerator<any, void, unknown>;
+    /**
+     * @deprecated This method uses the deprecated search_parameters API.
+     *
+     * Migrate to using chat() with enable_web_search instead:
+     *
+     * ```typescript
+     * const searchOptions: SearchOptions = {
+     *   enable_web_search: true,
+     *   web_search_filters: {
+     *     allowed_domains: ['example.com'],  // optional
+     *     excluded_domains: ['spam.com'],    // optional
+     *     enable_image_understanding: true   // optional
+     *   }
+     * };
+     *
+     * await client.chat([{role: "user", content: query}], [], undefined, searchOptions);
+     * ```
+     *
+     * The search_parameters field was deprecated by Grok API on December 15, 2025.
+     * See: https://docs.x.ai/docs/guides/tools/search-tools
+     */
     search(query: string, searchParameters?: SearchParameters): Promise<LLMResponse>;
     /**
      * Returns true if the host part of baseURL matches one of the allowedHosts or a subdomain thereof.