Your Name
01aa7f3e8f
gates / consistency-and-conventions (push) Successful in 32s
User hit 'API error 402: ... You requested up to 131072 tokens, but can only afford 4511' on the assist alias: no provider ever sent max_tokens, so OpenRouter's credit pre-check billed the routed model's full worst-case output; user also asked to bound session history to the last 5 requests/responses. Architect decisions: - AI_MAX_TOKENS (num, default 2048): sent as max_tokens on OpenRouter and generationConfig.maxOutputTokens on Gemini — a real per-request cost ceiling. llamacpp unchanged (local/free, no pre-check). - AI_SESSION_TURNS (num, default 40 kept back-compat; messages, 2 per exchange — 10 = last 5 conversations): resolved lazily in session_push because config loads after the hardcoded line-25 default. - Both registered in the bin/pos-ai POS_CONFIG @General section, so they appear in 'pos config ai' with num: validation. Reviewer hardening (CHANGES_REQUIRED -> fixed): unguarded env input could reach jq tonumber (0/-5/010/abc all savable via config-ui's ^-?[0-9]+$) and abort the CLI; both providers and session_push now guard with ^[1-9][0-9]*$ and fall back to the default. Verified: fake-curl shim smoke (16 provider-body + 12 session-window checks incl. the 010-regression proof), make gen idempotent, make check OK, make lint 0 FAIL/0 WARN, make test 17 files / 299 checks / 0 fail (~49s), bash -n clean, git diff --check clean. Reviewer ACCEPT (twice). Tester regression round (permanent provider-body + session-pruning coverage) intentionally not run this cycle — user's call; remains a documented follow-up.
710 lines
29 KiB
Bash
Executable File
710 lines
29 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
set -euo pipefail
|
|
# POS: ai ask — AI assistant: ask, chat, sessions, capture, models, providers
|
|
# POS_SUBCMDS: ask chat sessions capture models providers llamacpp
|
|
# POS_FLAGS: --provider --model --session --system --full --last --trust --no-command-execution
|
|
# POS_CONFIG: ai | ai.env | AI_PROVIDER=:Provider (gemini, openrouter or llamacpp, default gemini) | @[AI_PROVIDER=gemini|] Gemini | *providers=gemini | @[AI_PROVIDER=openrouter] OpenRouter | *providers=openrouter | @[AI_PROVIDER=llamacpp] llamacpp | LLAMACPP_PORT=:Server port (default 8088) | LLAMACPP_HOST=:Bind address (default 127.0.0.1) | LLAMACPP_MODEL=:Default model path (GGUF) | LLAMACPP_CTX_SIZE=num:Context window size (default 4096) | LLAMACPP_GPU_LAYERS=num:GPU layers (-1=auto, 0=CPU, default -1) | LLAMACPP_THREADS=num:CPU threads (default: nproc) | *providers=llamacpp | @General | AI_SYSTEM_PROMPT=:Custom system prompt (overrides built-in, empty to reset) | AI_MAX_TOKENS=num:Max output tokens per request (default 2048; OpenRouter/Gemini cost cap) | AI_SESSION_TURNS=num:Session message cap — 2 per exchange (default 40 = 20 exchanges; 10 = last 5)
|
|
|
|
source "$(dirname "$0")/../lib/common.sh" 2>/dev/null || source "$(dirname "$0")/common.sh"
|
|
|
|
# Shared config loader (canonical KEY=VALUE parser, env-wins precedence)
|
|
source "$(dirname "$0")/../lib/config-ui.sh" 2>/dev/null || source "$(dirname "$0")/config-ui.sh"
|
|
|
|
# ── Paths & constants ──────────────────────────────────────────
|
|
CONFIG_FILE="${CONFIG_FILE:-$CONFIG_DIR/ai.env}"
|
|
SESSION_DIR="$HOME/.local/share/linux_post_install/ai"
|
|
DISPATCH_LOG_DIR="$HOME/.local/share/linux_post_install/logs" # bin/pos per-run logs
|
|
LAST_CMD_OUTPUT_FILE="$HOME/.local/share/linux_post_install/last_cmd_output" # --last fallback for any command
|
|
OS_RELEASE_FILE="${OS_RELEASE_FILE:-/etc/os-release}" # read-only test seam (DEV.md env-overridable paths)
|
|
PROVIDER_DIR="$(dirname "$0")/../lib/ai-providers"
|
|
# Fallback for installed layout (flat /usr/local/bin)
|
|
[ -d "$PROVIDER_DIR" ] || PROVIDER_DIR="$(dirname "$0")/ai-providers"
|
|
|
|
SESSION="default"
|
|
SYSTEM_PROMPT=""
|
|
MAX_SESSION_TURNS=40
|
|
LAST_LOG_MAX_BYTES=4096
|
|
LAST_LOG_STALE_SECS=3600 # --last: warn when the attached log is older than this
|
|
# Built-in terse ask prompt. cmd_ask appends a machine-context clause
|
|
# (see machine_context) on EVERY ask — even --system/--full.
|
|
DEFAULT_SYSTEM_PROMPT_HARD="You are a Linux CLI assistant. Rules:
|
|
1. Lead with exact command(s) — no explanations unless asked
|
|
2. One line max per command; short bullets for multi-step only
|
|
3. No greetings, no pleasantries, no closing offers
|
|
4. For errors: diagnose and give the fix command first
|
|
5. Match the user's OS/package manager (apt/dnf/pacman)"
|
|
|
|
# Legacy: kept for session migration and backward compat config
|
|
LEGACY_GEMINI_CONFIG="${LEGACY_GEMINI_CONFIG:-$CONFIG_DIR/ai.env}"
|
|
LEGACY_OPENROUTER_CONFIG="${LEGACY_OPENROUTER_CONFIG:-$CONFIG_DIR/ai-openrouter.env}"
|
|
|
|
usage() {
|
|
cat <<EOF
|
|
Usage: pos ai [subcommand] [--provider <name>] [--model <id>] [--session <name>] [--system <text>] [--full] [--last] [--trust] [--no-command-execution]
|
|
|
|
AI assistant with pluggable providers (gemini, openrouter, llamacpp).
|
|
|
|
Subcommands:
|
|
ask "<prompt>" Answer; prints the answer text to stdout. The prompt may
|
|
also be piped in via stdin when no argument is given.
|
|
Runs inside the persistent 'default' session (prior turns
|
|
are sent as context); --session <name> picks another.
|
|
capture <cmd..> Run a command, capture its output for --last, and show it.
|
|
Each capture overwrites the previous one (latest only).
|
|
chat Interactive multi-turn conversation (session 'default'
|
|
unless --session is given).
|
|
models List available models for the active provider.
|
|
providers List available providers and their config status.
|
|
sessions List persistent sessions / clear one:
|
|
'sessions' and 'sessions reset <name>'.
|
|
|
|
Options:
|
|
--provider <name> Provider to use (gemini|openrouter|llamacpp; default: gemini).
|
|
Can also be set via AI_PROVIDER env/config.
|
|
--model <id> Override the model for this invocation.
|
|
--session <name> Use a named persistent session instead of 'default':
|
|
~/.local/share/linux_post_install/ai/<name>.json
|
|
(capped at $MAX_SESSION_TURNS turns).
|
|
--system <text> System instruction sent with every turn (kept out of the
|
|
session file); replaces the built-in terse ask prompt
|
|
wholesale, e.g. "Reply like a friendly Telegram chat".
|
|
--full Skip the built-in terse prompt — long-form answers.
|
|
--last ask only: attach the most recent pos dispatcher log or
|
|
captured output (tail, max $LAST_LOG_MAX_BYTES chars) so
|
|
the model can diagnose a real failure. Sources in priority
|
|
order: (1) newest pos log, (2) captured output from
|
|
'capture'. Notes on stderr which source was attached and
|
|
its age; warns when stale (>60 min).
|
|
--trust Auto-execute agent-detected commands without confirmation.
|
|
Interactive terminals only — it has no effect when stdout/
|
|
stdin is not a tty (e.g. chat bridges, cron), which never
|
|
execute commands. Used by trusted alias wrappers — do NOT
|
|
pass manually unless you fully trust the agent's output.
|
|
--no-command-execution Never run agent-detected command blocks. The proposed
|
|
command is neither printed nor executed; no
|
|
confirmation prompt is shown (structural guard — used by
|
|
chat bridges so a future refactor cannot auto-execute).
|
|
Each of these two modes contradict the other: the last one
|
|
given on the command line wins.
|
|
-h|--help This help.
|
|
|
|
Config: $CONFIG_FILE (edit with 'pos config ai')
|
|
AI_PROVIDER Provider to use (gemini|openrouter|llamacpp, default gemini)
|
|
AI_SYSTEM_PROMPT Custom system prompt (overrides built-in; empty to reset)
|
|
AI_MAX_TOKENS Max output tokens per request (default 2048; OpenRouter/Gemini cost cap)
|
|
AI_SESSION_TURNS Session message cap — 2 per exchange (default 40 = 20 exchanges; 10 = last 5)
|
|
Provider keys: auto-discovered from lib/ai-providers/*.sh
|
|
(AI_GEMINI_API_KEY, OPENROUTER_API_KEY, etc.)
|
|
|
|
Notes:
|
|
ask is terse by default: a built-in system instruction tells the model to
|
|
lead with the exact commands and keep prose minimal — and to diagnose pasted
|
|
errors/output with a fix first. That prompt ends with one machine-context
|
|
line (hostname, distro, kernel, arch detected on this box) so answers fit
|
|
the actual machine; --system replaces it wholesale, --full drops it all.
|
|
Every ask/chat lands in a persistent session ('default' unless --session);
|
|
clear it with 'pos ai sessions reset default'. On a terminal the
|
|
answer is rendered as markdown (glow if installed, else a built-in
|
|
renderer); when stdout is not a tty (pipes, scripts, Telegram bridges) the
|
|
raw markdown is printed unchanged.
|
|
|
|
Examples:
|
|
pos ai ask "check disk space on /"
|
|
pos ai --provider gemini ask "Explain DNS in one line"
|
|
pos ai --provider openrouter ask "hi"
|
|
echo "summarize this log" | pos ai ask
|
|
failing-cmd 2>&1 | pos ai ask how do I fix this
|
|
pos ai ask --last "why did that fail?" # attach last output
|
|
pos ai capture pip install xyz # capture any command
|
|
pos ai ask --last "what happened?" # after capture
|
|
pos ai chat
|
|
pos ai models
|
|
pos ai providers
|
|
pos ai ask --model gemini-2.5-flash "hi"
|
|
pos ai ask --system "Reply like a pirate" "explain chmod"
|
|
pos ai ask --session work "my name is joe"
|
|
pos ai ask --session work "what is my name?" # remembers
|
|
pos ai sessions
|
|
pos ai sessions reset default # forget default memory
|
|
EOF
|
|
exit 0
|
|
}
|
|
|
|
# ── Provider loading ───────────────────────────────────────────
|
|
load_provider() {
|
|
local p="${PROVIDER:-gemini}"
|
|
local f="$PROVIDER_DIR/$p.sh"
|
|
[ -f "$f" ] || err "Unknown provider '$p' — available: $(ls "$PROVIDER_DIR"/*.sh 2>/dev/null | xargs -I{} basename {} .sh | tr '\n' ' ')"
|
|
# shellcheck source=/dev/null
|
|
source "$f"
|
|
}
|
|
|
|
# ── ai.env loader (canonical loader + legacy files) ─────────────
|
|
load_config() {
|
|
load_env_file "$CONFIG_FILE"
|
|
# Legacy provider-specific config files (fallback for old configs)
|
|
# Both files are loaded — env-var precedence means unified AI_API_KEY wins.
|
|
local legacy_files="$LEGACY_GEMINI_CONFIG $LEGACY_OPENROUTER_CONFIG"
|
|
local legacy_env
|
|
for legacy_env in $legacy_files; do
|
|
[ -f "$legacy_env" ] && [ "$legacy_env" != "$CONFIG_FILE" ] || continue
|
|
load_env_file "$legacy_env"
|
|
done
|
|
}
|
|
|
|
# ── Config resolution ──────────────────────────────────────────
|
|
resolve_key() {
|
|
load_config
|
|
local p="${PROVIDER:-gemini}"
|
|
# Each provider has its own API key — set AI_API_KEY internally for adapters
|
|
case "$p" in
|
|
gemini) [ -n "${AI_GEMINI_API_KEY:-}" ] && export AI_API_KEY="$AI_GEMINI_API_KEY" && return 0 ;;
|
|
openrouter) [ -n "${OPENROUTER_API_KEY:-}" ] && export AI_API_KEY="$OPENROUTER_API_KEY" && return 0 ;;
|
|
llamacpp) return 0 ;; # No API key needed for local server
|
|
esac
|
|
return 1
|
|
}
|
|
|
|
require_key() {
|
|
if ! resolve_key >/dev/null 2>&1; then
|
|
local p="${PROVIDER:-gemini}"
|
|
case "$p" in
|
|
gemini) err "No Gemini API key — run 'pos config ai' and set AI_GEMINI_API_KEY" ;;
|
|
openrouter) err "No OpenRouter API key — run 'pos config ai' and set OPENROUTER_API_KEY" ;;
|
|
llamacpp) ;; # No key needed for local server
|
|
esac
|
|
err "No API key for provider '$p' — run 'pos config ai'"
|
|
fi
|
|
}
|
|
|
|
resolve_model() {
|
|
local p="${PROVIDER:-gemini}"
|
|
if [ -n "${MODEL_OVERRIDE:-}" ]; then
|
|
printf '%s' "$MODEL_OVERRIDE"
|
|
elif [ -n "${AI_MODEL:-}" ]; then
|
|
printf '%s' "$AI_MODEL"
|
|
else
|
|
# Provider-specific fallback
|
|
case "$p" in
|
|
gemini) [ -n "${AI_GEMINI_MODEL:-}" ] && printf '%s' "$AI_GEMINI_MODEL" && return ;;
|
|
openrouter) [ -n "${OPENROUTER_MODEL:-}" ] && printf '%s' "$OPENROUTER_MODEL" && return ;;
|
|
llamacpp) [ -n "${LLAMACPP_MODEL:-}" ] && printf '%s' "$(basename "$LLAMACPP_MODEL")" && return ;;
|
|
esac
|
|
provider_default_model
|
|
fi
|
|
}
|
|
|
|
# ── --last: attach the most recent pos command output ───────────
|
|
newest_pos_log() {
|
|
[ -d "$DISPATCH_LOG_DIR" ] || return 1
|
|
local f
|
|
while IFS= read -r f; do
|
|
[ -s "$f" ] && { printf '%s' "$f"; return 0; }
|
|
done < <(ls "$DISPATCH_LOG_DIR"/*.log 2>/dev/null | LC_ALL=C sort -r | grep -v '/pos\.log$')
|
|
return 1
|
|
}
|
|
|
|
last_log_context() {
|
|
local raw
|
|
raw="$(tail -c "$LAST_LOG_MAX_BYTES" "$1")"
|
|
if command -v iconv >/dev/null 2>&1; then
|
|
raw="$(printf '%s' "$raw" | iconv -c -f utf-8 -t utf-8 2>/dev/null || printf '%s' "$raw")"
|
|
fi
|
|
if [ "$(wc -c <"$1")" -gt "$LAST_LOG_MAX_BYTES" ]; then
|
|
printf '[…truncated…]\n%s' "$raw"
|
|
else
|
|
printf '%s' "$raw"
|
|
fi
|
|
}
|
|
|
|
human_age() {
|
|
local s="$1"
|
|
[ "$s" -lt 0 ] && s=0
|
|
if [ "$s" -lt 60 ]; then printf 'just now'
|
|
elif [ "$s" -lt 3600 ]; then printf '%sm' "$((s / 60))"
|
|
elif [ "$s" -lt 86400 ]; then printf '%sh' "$((s / 3600))"
|
|
else printf '%sd' "$((s / 86400))"
|
|
fi
|
|
}
|
|
|
|
last_log_annotate() {
|
|
local f="$1" age_s age line
|
|
age_s=$(( $(date +%s) - $(stat -c %Y "$f") ))
|
|
[ "$age_s" -lt 0 ] && age_s=0
|
|
age="$(human_age "$age_s")"
|
|
printf '[i] attaching last pos output — %s (%s)\n' "$(basename "$f")" "$age" >&2
|
|
line="$(grep -m1 '[^[:space:]]' "$f" 2>/dev/null || true)"
|
|
if [ -n "$line" ]; then
|
|
printf '[i] "%.100s"\n' "$line" >&2
|
|
fi
|
|
if [ "$age_s" -gt "$LAST_LOG_STALE_SECS" ]; then
|
|
printf '[!] that log is %s old and may not match your current problem. For a FRESH failure of any command: failing-cmd 2>&1 | pos ai ask "what happened"\n' "$age" >&2
|
|
fi
|
|
}
|
|
|
|
# ── Persistent session memory (universal OpenAI messages format) ─
|
|
session_file() {
|
|
local name="${1:-$SESSION}"
|
|
name="${name//[^A-Za-z0-9_-]/_}"
|
|
printf '%s/%s.json' "$SESSION_DIR" "$name"
|
|
}
|
|
|
|
session_load() {
|
|
[ -n "$SESSION" ] || { printf '{"messages":[]}'; return 0; }
|
|
local f
|
|
f="$(session_file)"
|
|
if [ -s "$f" ]; then
|
|
# Check for old Gemini contents format → migrate transparently
|
|
if jq -e '.contents' "$f" >/dev/null 2>&1 && ! jq -e '.messages' "$f" >/dev/null 2>&1; then
|
|
local migrated
|
|
migrated="$(jq -c '{messages: [.contents[]? | {role: (if .role == "model" then "assistant" else .role end), content: (.parts | map(.text) | join(""))}]}' "$f" 2>/dev/null)" || {
|
|
printf '{"messages":[]}'; return 0
|
|
}
|
|
printf '%s\n' "$migrated" > "$f"
|
|
chmod 600 "$f"
|
|
printf '%s' "$migrated"
|
|
elif jq -e '.messages' "$f" >/dev/null 2>&1; then
|
|
cat "$f"
|
|
else
|
|
printf '{"messages":[]}'
|
|
fi
|
|
else
|
|
printf '{"messages":[]}'
|
|
fi
|
|
}
|
|
|
|
session_save() {
|
|
[ -n "$SESSION" ] || return 0
|
|
local f tmp
|
|
f="$(session_file)"
|
|
mkdir -p "$SESSION_DIR"
|
|
tmp="$(mktemp)"
|
|
printf '%s\n' "$1" >"$tmp"
|
|
mv "$tmp" "$f"
|
|
chmod 600 "$f"
|
|
}
|
|
|
|
# Append a turn and prune to the last MAX_SESSION_TURNS entries. stdout = JSON.
|
|
session_push() {
|
|
local messages="$1" role="$2" text="$3"
|
|
local n="${AI_SESSION_TURNS:-$MAX_SESSION_TURNS}"
|
|
[[ "$n" =~ ^[1-9][0-9]*$ ]] || n="$MAX_SESSION_TURNS"
|
|
printf '%s' "$messages" | jq -c --arg r "$role" --arg t "$text" --argjson n "$n" \
|
|
'.messages += [{role:$r, content:$t}] | .messages |= .[-$n:]'
|
|
}
|
|
|
|
# ── Terminal markdown rendering (tty-only; raw bytes otherwise) ──
|
|
render_markdown() {
|
|
local text="$1"
|
|
# Check stdout tty OR controlling terminal (/dev/tty) — the shell hook
|
|
# (pos-ai-hook.sh) redirects stdout through tee, breaking [ -t 1 ], but
|
|
# /dev/tty remains writable in interactive shells.
|
|
if [ ! -t 1 ] && [ ! -w /dev/tty ]; then
|
|
printf '%s\n' "$text"
|
|
return 0
|
|
fi
|
|
local rendered prog
|
|
prog='
|
|
BEGIN {
|
|
e = sprintf("%c", 27)
|
|
R = e "[0m"; DIM = e "[2m"; B = e "[1m"
|
|
YEL = e "[33m"; CYA = e "[1;36m"
|
|
RULE = ""
|
|
for (i = 0; i < 60; i++) RULE = RULE "─"
|
|
RULE = DIM RULE R
|
|
}
|
|
/^```/ { fence = !fence; next }
|
|
fence { printf "%s %s%s\n", DIM, $0, R; next }
|
|
/^#{1,4} / {
|
|
sub(/^#{1,4} +/, "")
|
|
printf "%s%s%s\n", CYA, $0, R
|
|
next
|
|
}
|
|
/^(-{3,}|\*{3,}|_{3,})$/ { print RULE; next }
|
|
{
|
|
line = $0
|
|
out = ""; rest = line
|
|
while (match(rest, /`[^`]*`/)) {
|
|
out = out substr(rest, 1, RSTART - 1) YEL \
|
|
substr(rest, RSTART + 1, RLENGTH - 2) R
|
|
rest = substr(rest, RSTART + RLENGTH)
|
|
}
|
|
line = out rest
|
|
out = ""; rest = line
|
|
while (match(rest, /\*\*[^*]+\*\*/)) {
|
|
out = out substr(rest, 1, RSTART - 1) B \
|
|
substr(rest, RSTART + 2, RLENGTH - 4) R
|
|
rest = substr(rest, RSTART + RLENGTH)
|
|
}
|
|
line = out rest
|
|
out = ""; rest = line
|
|
while (match(rest, /__[^_]+__/)) {
|
|
out = out substr(rest, 1, RSTART - 1) B \
|
|
substr(rest, RSTART + 2, RLENGTH - 4) R
|
|
rest = substr(rest, RSTART + RLENGTH)
|
|
}
|
|
print out rest
|
|
}
|
|
'
|
|
if command -v glow >/dev/null 2>&1; then
|
|
rendered="$(printf '%s\n' "$text" | glow -)"
|
|
else
|
|
rendered="$(printf '%s\n' "$text" | awk "$prog")"
|
|
fi
|
|
printf '\n%s\n' "$rendered"
|
|
}
|
|
|
|
# ── Command extraction from AI responses ────────────────────────
|
|
_extract_commands() {
|
|
local text="$1"
|
|
printf '%s' "$text" | awk '
|
|
/^```(bash|sh|shell)/ { in_block=1; next }
|
|
/^```/ { if (in_block) in_block=0; next }
|
|
in_block && NF > 0 { lines[++n] = $0 }
|
|
END {
|
|
for (i = 1; i <= n; i++) {
|
|
if (i > 1) printf "\n"
|
|
printf "%s", lines[i]
|
|
}
|
|
}
|
|
'
|
|
}
|
|
|
|
# ── Interactive prompt to run extracted commands ─────────────────
|
|
_prompt_run_command() {
|
|
local cmd="$1" trusted="${2:-0}"
|
|
# Structural guards: commands never execute without a controlling tty, and
|
|
# never when --no-command-execution was passed. Both apply BEFORE any
|
|
# auto-execute path (including --trust), so a future refactor cannot
|
|
# introduce a non-interactive or bridge-forced execution.
|
|
[ -w /dev/tty ] || return 0
|
|
if [ "${NO_EXEC:-0}" -eq 1 ]; then
|
|
return 0
|
|
fi
|
|
printf '\n%s\n' "Command detected:" >&2
|
|
printf ' %s\n\n' "$cmd" >&2
|
|
if [ "$trusted" -eq 1 ]; then
|
|
printf '[trusted] Auto-executing (no confirmation)\n\n' >&2
|
|
printf '%s\n' "$cmd"
|
|
run eval "$cmd"
|
|
return
|
|
fi
|
|
printf 'Run this command? [y/N] ' >&2
|
|
local choice
|
|
IFS= read -r choice </dev/tty || choice=""
|
|
case "$choice" in
|
|
[yY])
|
|
printf '%s\n' "$cmd"
|
|
run eval "$cmd"
|
|
;;
|
|
*)
|
|
# Add to shell history so user can press ↑ to recall, edit, run
|
|
history -s "$cmd" 2>/dev/null || true
|
|
printf '%s\n' "Command added to history — press ↑ to recall, edit, and run." >&2
|
|
;;
|
|
esac
|
|
}
|
|
|
|
# ── Machine context appended to the built-in default prompt ─────
|
|
mc_clean() {
|
|
sed -e 's/\x1b\[[0-9;]*[A-Za-z]//g' \
|
|
-e 's/[[:space:]][[:space:]]*/ /g' \
|
|
| tr -d '\000-\010\013-\037\177' \
|
|
| sed -e 's/^ //; s/ $//'
|
|
}
|
|
|
|
machine_context() {
|
|
local raw line key val h="" o="" k="" a="" part out=""
|
|
if command -v hostnamectl >/dev/null 2>&1; then
|
|
raw="$(hostnamectl status 2>/dev/null || true)"
|
|
while IFS= read -r line; do
|
|
key="$(printf '%s' "${line%%:*}" | tr -d '[:space:]')"
|
|
val="${line#*:}"
|
|
case "$key" in
|
|
Statichostname|Transienthostname|Hostname)
|
|
[ -z "$h" ] && h="$val" ;;
|
|
OperatingSystem)
|
|
[ -z "$o" ] && o="$val" ;;
|
|
Kernel)
|
|
[ -z "$k" ] && k="$val" ;;
|
|
Architecture)
|
|
[ -z "$a" ] && a="$val" ;;
|
|
esac
|
|
done <<< "$raw"
|
|
fi
|
|
if [ -z "$o" ] && [ -r "$OS_RELEASE_FILE" ]; then
|
|
o="$(
|
|
. "$OS_RELEASE_FILE" 2>/dev/null || true
|
|
if [ -n "${PRETTY_NAME:-}" ]; then
|
|
printf '%s' "$PRETTY_NAME"
|
|
elif [ -n "${NAME:-}" ]; then
|
|
printf '%s' "${NAME}${VERSION_ID:+ (${VERSION_ID})}"
|
|
fi
|
|
)"
|
|
fi
|
|
[ -n "$k" ] || k="$(uname -sr 2>/dev/null || true)"
|
|
[ -n "$a" ] || a="$(uname -m 2>/dev/null || true)"
|
|
h="$(printf '%s' "$h" | mc_clean)"
|
|
o="$(printf '%s' "$o" | mc_clean)"
|
|
k="$(printf '%s' "$k" | mc_clean)"
|
|
a="$(printf '%s' "$a" | mc_clean)"
|
|
case "$k" in "Linux "*) k="${k#Linux }" ;; esac
|
|
local out=""
|
|
for part in "$h" "$o" "${k:+kernel $k}" "$a"; do
|
|
[ -n "$part" ] || continue
|
|
if [ -n "$out" ]; then out="$out, $part"; else out="$part"; fi
|
|
done
|
|
[ -n "$out" ] || return 0
|
|
printf 'Machine context (answers must fit this box): %s.' "$out"
|
|
}
|
|
|
|
# ── Subcommands ────────────────────────────────────────────────
|
|
cmd_capture() {
|
|
[ $# -gt 0 ] || err "usage: pos ai capture <command> [args...]"
|
|
mkdir -p "$(dirname "$LAST_CMD_OUTPUT_FILE")"
|
|
"$@" 2>&1 | tee "$LAST_CMD_OUTPUT_FILE"
|
|
local rc=${PIPESTATUS[0]}
|
|
printf '[captured → %s]\n' "$LAST_CMD_OUTPUT_FILE" >&2
|
|
return $rc
|
|
}
|
|
|
|
cmd_ask() {
|
|
local prompt="" messages out system ctx mc
|
|
if [ $# -gt 0 ]; then
|
|
prompt="$*"
|
|
elif [ ! -t 0 ]; then
|
|
prompt="$(cat)"
|
|
fi
|
|
[ -n "$prompt" ] || err "No prompt given — usage: pos ai ask \"<prompt>\""
|
|
# --last: append the most recent logged pos command output AFTER the
|
|
# question, so the model diagnoses the real failure.
|
|
if [ "$LAST_MODE" -eq 1 ]; then
|
|
local log_file="" pos_log="" captured_log=""
|
|
pos_log="$(newest_pos_log 2>/dev/null)" || true
|
|
[ -s "$LAST_CMD_OUTPUT_FILE" ] && captured_log="$LAST_CMD_OUTPUT_FILE"
|
|
if [ -n "$pos_log" ] && [ -n "$captured_log" ]; then
|
|
local pos_age=$(( $(date +%s) - $(stat -c %Y "$pos_log") ))
|
|
local cap_age=$(( $(date +%s) - $(stat -c %Y "$captured_log") ))
|
|
if [ "$cap_age" -lt "$pos_age" ]; then
|
|
log_file="$captured_log"
|
|
else
|
|
log_file="$pos_log"
|
|
fi
|
|
elif [ -n "$captured_log" ]; then
|
|
log_file="$captured_log"
|
|
else
|
|
log_file="$pos_log"
|
|
fi
|
|
[ -n "$log_file" ] || err "No recent output found — run 'pos ai capture <cmd>' first, or pipe: cmd 2>&1 | pos ai ask \"what happened\""
|
|
last_log_annotate "$log_file"
|
|
ctx="$(last_log_context "$log_file")"
|
|
prompt="$prompt"$'\n\n[last command output:]\n'"$ctx"
|
|
fi
|
|
require_key
|
|
# Terse by default: user --system replaces the built-in prompt wholesale;
|
|
# --full skips the built-in text. The machine-context line (hostname,
|
|
# distro, kernel, arch) is ALWAYS appended so every ask knows the box.
|
|
system="$SYSTEM_PROMPT"
|
|
if [ -z "$system" ] && [ "$FULL_MODE" -eq 0 ]; then
|
|
# Check AI_SYSTEM_PROMPT config first, then fall back to built-in
|
|
system="${AI_SYSTEM_PROMPT:-}"
|
|
if [ -z "$system" ]; then
|
|
system="$DEFAULT_SYSTEM_PROMPT_HARD"
|
|
fi
|
|
fi
|
|
mc="$(machine_context)"
|
|
if [ -n "$system" ]; then
|
|
[ -n "$mc" ] && mc=" $mc"
|
|
system="$system$mc"
|
|
else
|
|
# --full (no system text): carry machine context as the sole system turn
|
|
system="$mc"
|
|
fi
|
|
# Persistent session memory ('default' unless --session).
|
|
messages="$(session_load)"
|
|
messages="$(session_push "$messages" user "$prompt")"
|
|
if ! out="$(provider_generate "$(resolve_model)" "$messages" "$system" 2>&1)"; then
|
|
err "$out"
|
|
fi
|
|
messages="$(session_push "$messages" assistant "$out")"
|
|
session_save "$messages"
|
|
render_markdown "$out"
|
|
# Command execution prompt: extract commands from response and offer to run
|
|
local _cmd
|
|
_cmd="$(_extract_commands "$out")"
|
|
[ -n "$_cmd" ] && _prompt_run_command "$_cmd" "$TRUST_MODE"
|
|
}
|
|
|
|
cmd_chat() {
|
|
[ $# -eq 0 ] || err "Unexpected argument for chat: $*"
|
|
local model messages text answer provider_display
|
|
model="$(resolve_model)"
|
|
require_key
|
|
provider_display="$(provider_name)"
|
|
messages="$(session_load)"
|
|
printf 'session: %s (resumed %s prior turns)\n' "$SESSION" "$(printf '%s' "$messages" | jq -r '.messages | length')"
|
|
trap 'echo; echo "bye"; exit 0' INT
|
|
echo "${provider_display} · ${model} — type a message; q=quit, /reset=clear history"
|
|
while true; do
|
|
printf '> '
|
|
IFS= read -r text || break
|
|
case "$text" in
|
|
"" ) continue ;;
|
|
q|Q|quit|exit) echo; echo "bye"; return 0 ;;
|
|
/reset)
|
|
messages='{"messages":[]}'
|
|
session_save "$messages"
|
|
echo "[history cleared]"
|
|
continue ;;
|
|
esac
|
|
messages="$(session_push "$messages" user "$text")"
|
|
if ! answer="$(provider_generate "$model" "$messages" "$SYSTEM_PROMPT" 2>&1)"; then
|
|
warn "AI error: $answer"
|
|
continue
|
|
fi
|
|
messages="$(session_push "$messages" assistant "$answer")"
|
|
session_save "$messages"
|
|
printf '\n'
|
|
render_markdown "$answer"
|
|
# Command execution prompt: extract commands from response and offer to run
|
|
local _cmd
|
|
_cmd="$(_extract_commands "$answer")"
|
|
[ -n "$_cmd" ] && _prompt_run_command "$_cmd" "$TRUST_MODE"
|
|
printf '\n\n'
|
|
done
|
|
echo
|
|
return 0
|
|
}
|
|
|
|
cmd_sessions() {
|
|
local action="${1:-list}" name f n
|
|
case "$action" in
|
|
list|"")
|
|
[ -d "$SESSION_DIR" ] || { echo "no sessions"; return 0; }
|
|
local found=0
|
|
for f in "$SESSION_DIR"/*.json; do
|
|
[ -f "$f" ] || continue
|
|
found=1
|
|
n="$(jq -r '.messages | length' "$f" 2>/dev/null || echo 0)"
|
|
printf ' %-32s %s turns\n' "$(basename "$f" .json)" "${n:-0}"
|
|
done
|
|
[ "$found" -eq 1 ] || echo "no sessions"
|
|
;;
|
|
reset)
|
|
[ $# -ge 2 ] || err "usage: pos ai sessions reset <name>"
|
|
name="$2"
|
|
if rm -f "$(session_file "$name")"; then
|
|
ok "session '$name' cleared"
|
|
fi
|
|
;;
|
|
*) err "Unknown sessions subcommand '$action' (list | reset <name>)" ;;
|
|
esac
|
|
}
|
|
|
|
cmd_models() {
|
|
[ $# -eq 0 ] || err "Unexpected argument for models: $*"
|
|
local model
|
|
model="$(resolve_model)"
|
|
require_key
|
|
provider_models_list "$model"
|
|
}
|
|
|
|
cmd_providers() {
|
|
echo "Available providers:"
|
|
load_config # ensure env vars are populated
|
|
local active="${PROVIDER:-gemini}"
|
|
for f in "$PROVIDER_DIR"/*.sh; do
|
|
[ -f "$f" ] || continue
|
|
local name pname pmodel configured current
|
|
name="$(basename "$f" .sh)"
|
|
# Source provider in a subshell to get its metadata
|
|
local meta
|
|
meta="$( ( source "$f"; printf '%s\x00%s' "$(provider_name)" "$(provider_default_model)" ) 2>/dev/null )" || true
|
|
pname="${meta%%$'\x00'*}"
|
|
pmodel="${meta#*$'\x00'}"
|
|
[ -n "$pname" ] || pname="$name"
|
|
[ -n "$pmodel" ] || pmodel="unknown"
|
|
# Check if API key exists for this provider
|
|
configured="not configured"
|
|
case "$name" in
|
|
gemini) [ -n "${AI_GEMINI_API_KEY:-}" ] && configured="configured" ;;
|
|
openrouter) [ -n "${OPENROUTER_API_KEY:-}" ] && configured="configured" ;;
|
|
llamacpp) configured="configured" ;; # Local server — always configured
|
|
esac
|
|
current=""
|
|
[ "$name" = "$active" ] && current=" ← active"
|
|
printf ' %-16s %s (model: %s)%s\n' "$name" "$configured" "$pmodel" "$current"
|
|
done
|
|
}
|
|
|
|
# ── Parse flags + subcommand ────────────────────────────────────
|
|
MODEL_OVERRIDE=""
|
|
FULL_MODE=0
|
|
LAST_MODE=0
|
|
TRUST_MODE=0
|
|
NO_EXEC=0
|
|
PROVIDER=""
|
|
cmd=""
|
|
args=()
|
|
while [ $# -gt 0 ]; do
|
|
case "$1" in
|
|
-h|--help) usage ;;
|
|
--provider)
|
|
[ $# -ge 2 ] || err "--provider needs a value"
|
|
PROVIDER="$2"; shift 2 ;;
|
|
--model)
|
|
[ $# -ge 2 ] || err "--model needs a value"
|
|
MODEL_OVERRIDE="$2"; shift 2 ;;
|
|
--session)
|
|
[ $# -ge 2 ] || err "--session needs a value"
|
|
SESSION="$2"; shift 2 ;;
|
|
--system)
|
|
[ $# -ge 2 ] || err "--system needs a value"
|
|
SYSTEM_PROMPT="$2"; shift 2 ;;
|
|
--full)
|
|
FULL_MODE=1; shift ;;
|
|
--last)
|
|
LAST_MODE=1; shift ;;
|
|
--trust)
|
|
TRUST_MODE=1; NO_EXEC=0; shift ;;
|
|
--no-command-execution)
|
|
NO_EXEC=1; TRUST_MODE=0; shift ;;
|
|
-*) err "Unknown option '$1' (see --help)" ;;
|
|
*)
|
|
if [ -z "$cmd" ]; then
|
|
cmd="$1"
|
|
else
|
|
args+=("$1")
|
|
fi
|
|
shift ;;
|
|
esac
|
|
done
|
|
|
|
# Resolve provider: --provider flag > AI_PROVIDER env/config > default gemini
|
|
if [ -z "$PROVIDER" ]; then
|
|
load_config
|
|
PROVIDER="${AI_PROVIDER:-gemini}"
|
|
fi
|
|
|
|
# Load provider adapter functions
|
|
load_provider
|
|
|
|
if [ "$LAST_MODE" -eq 1 ] && [ "${cmd:-}" != "ask" ]; then
|
|
err "--last only applies to 'ask' — for capturing output use 'capture': pos ai capture <cmd>"
|
|
fi
|
|
|
|
case "${cmd:-}" in
|
|
"") usage ;;
|
|
ask) cmd_ask "${args[@]}" ;;
|
|
capture) cmd_capture "${args[@]}" ;;
|
|
chat) cmd_chat "${args[@]}" ;;
|
|
models) cmd_models "${args[@]}" ;;
|
|
providers) cmd_providers "${args[@]}" ;;
|
|
sessions) cmd_sessions "${args[@]}" ;;
|
|
llamacpp)
|
|
# Provider shorthand (backward compat, same as the gemini/openrouter
|
|
# forwarders): pos ai llamacpp <subcmd> ... == pos ai --provider llamacpp <subcmd> ...
|
|
exec "$0" --provider llamacpp "${args[@]}" ;;
|
|
*) err "Unknown ai subcommand '$cmd' (see --help)" ;;
|
|
esac
|