perplexity-web-mcp-cli 0.12.0__tar.gz → 0.12.2__tar.gz

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: perplexity-web-mcp-cli
-Version: 0.12.0
+Version: 0.12.2
 Summary: CLI, MCP server, and Anthropic/OpenAI API-compatible interface for Perplexity AI.
 Keywords: perplexity,ai,mcp,anthropic,api,client
 Author: Jacob BD
@@ -41,6 +41,13 @@ Description-Content-Type: text/markdown
 
 # Perplexity Web MCP & CLI
 
+[![PyPI version](https://img.shields.io/pypi/v/perplexity-web-mcp-cli)](https://pypi.org/project/perplexity-web-mcp-cli/)
+[![PyPI downloads](https://img.shields.io/pypi/dm/perplexity-web-mcp-cli)](https://pypistats.org/packages/perplexity-web-mcp-cli)
+[![Total downloads](https://static.pepy.tech/badge/perplexity-web-mcp-cli)](https://pepy.tech/projects/perplexity-web-mcp-cli)
+[![Python](https://img.shields.io/pypi/pyversions/perplexity-web-mcp-cli)](https://pypi.org/project/perplexity-web-mcp-cli/)
+[![License](https://img.shields.io/pypi/l/perplexity-web-mcp-cli)](https://github.com/jacob-bd/perplexity-web-mcp/blob/main/LICENSE)
+[![Buy Me a Coffee](https://img.shields.io/badge/Buy%20Me%20a%20Coffee-FFDD00?style=flat-square&logo=buy-me-a-coffee&logoColor=black)](https://buymeacoffee.com/jacobbd)
+
 <p align="center">
   <a href="https://youtu.be/9xyClDvmoZ0">
     <img src="https://img.youtube.com/vi/9xyClDvmoZ0/maxresdefault.jpg" alt="Perplexity Powerhouse + Model Council Demo" width="400">
@@ -233,7 +240,7 @@ pwm research "NVIDIA competitive landscape" -s finance --json
 Query multiple models in parallel and get a synthesized consensus. Each model costs 1 Pro Search. Default synthesis uses Sonar 2 (also 1 Pro Search).
 
 ```bash
-# Default: GPT-5.4, Claude Opus, Gemini Pro + Sonar 2 synthesis (4 Pro Searches)
+# Default: GPT-5.4, Claude Sonnet, Gemini Pro + Sonar 2 synthesis (4 Pro Searches)
 pwm council "What are best practices for microservices?"
 ```
 

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/README.md

@@ -4,6 +4,13 @@
 
 # Perplexity Web MCP & CLI
 
+[![PyPI version](https://img.shields.io/pypi/v/perplexity-web-mcp-cli)](https://pypi.org/project/perplexity-web-mcp-cli/)
+[![PyPI downloads](https://img.shields.io/pypi/dm/perplexity-web-mcp-cli)](https://pypistats.org/packages/perplexity-web-mcp-cli)
+[![Total downloads](https://static.pepy.tech/badge/perplexity-web-mcp-cli)](https://pepy.tech/projects/perplexity-web-mcp-cli)
+[![Python](https://img.shields.io/pypi/pyversions/perplexity-web-mcp-cli)](https://pypi.org/project/perplexity-web-mcp-cli/)
+[![License](https://img.shields.io/pypi/l/perplexity-web-mcp-cli)](https://github.com/jacob-bd/perplexity-web-mcp/blob/main/LICENSE)
+[![Buy Me a Coffee](https://img.shields.io/badge/Buy%20Me%20a%20Coffee-FFDD00?style=flat-square&logo=buy-me-a-coffee&logoColor=black)](https://buymeacoffee.com/jacobbd)
+
 <p align="center">
   <a href="https://youtu.be/9xyClDvmoZ0">
     <img src="https://img.youtube.com/vi/9xyClDvmoZ0/maxresdefault.jpg" alt="Perplexity Powerhouse + Model Council Demo" width="400">
@@ -196,7 +203,7 @@ pwm research "NVIDIA competitive landscape" -s finance --json
 Query multiple models in parallel and get a synthesized consensus. Each model costs 1 Pro Search. Default synthesis uses Sonar 2 (also 1 Pro Search).
 
 ```bash
-# Default: GPT-5.4, Claude Opus, Gemini Pro + Sonar 2 synthesis (4 Pro Searches)
+# Default: GPT-5.4, Claude Sonnet, Gemini Pro + Sonar 2 synthesis (4 Pro Searches)
 pwm council "What are best practices for microservices?"
 ```
 

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "perplexity-web-mcp-cli"
-version = "0.12.0"
+version = "0.12.2"
 description = "CLI, MCP server, and Anthropic/OpenAI API-compatible interface for Perplexity AI."
 authors = [{ name = "Jacob BD" }]
 license = "MIT"

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/cli/ai_doc.py

@@ -66,7 +66,7 @@ MODEL COUNCIL
   pwm council "query" --json                  Output as JSON
 
   Each model in the council costs 1 Pro Search, plus 1 for synthesis. Default = 4 Pro Searches.
-  Available models: gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26
+  Available models: sonar, gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26
   Thinking toggle: -t / --thinking (gpt54, gpt55, claude_sonnet, claude_opus, kimi_k26 support toggle;
     gemini_pro and nemotron are always thinking)
 
@@ -110,7 +110,7 @@ MODELS
 Name            Identifier              Thinking   Notes
 -----------     ----------------------  ---------  ---------------------------
 auto            pplx_pro                No         Auto-selects best model
-sonar           experimental            No         Sonar 2 (latest in-house)
+sonar           experimental            No         Sonar 2 (concise search mode for grounded responses)
 deep_research   pplx_alpha              No         In-depth reports (monthly quota)
 gpt54           gpt54                   Yes        OpenAI GPT-5.4 (versatile)
 gpt55           gpt55                   Yes        OpenAI GPT-5.5 (latest, Max tier)
@@ -168,12 +168,13 @@ QUERY TOOLS (each call costs 1 Pro Search query unless noted):
   pplx_ask(query, source_focus="web")
       Auto-selects best model. 1 PRO SEARCH per call.
 
-  pplx_council(query, source_focus="web", models="gpt54,claude_opus,gemini_pro",
+  pplx_council(query, source_focus="web", models="gpt54,claude_sonnet,gemini_pro",
                synthesize=True, thinking=False, chairman="sonar")
       Model Council — N PRO SEARCHES (1 per model selected).
       BEFORE CALLING: You MUST ask the user which models and how many.
-      Available: gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26.
-      Default: 3 models (GPT-5.4, Claude Opus, Gemini Pro) + synthesis = 4 Pro Searches.
+      Available: sonar, gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26.
+      Max-only: gpt55, claude_opus. Exclude these when Subscription is Pro.
+      Default: 3 Pro-compatible models (GPT-5.4, Claude Sonnet, Gemini Pro) + synthesis = 4 Pro Searches.
       Synthesis uses Sonar 2 by default. Set chairman to override.
       Non-sonar chairman costs 1 extra Pro Search.
       Set synthesize=False to skip synthesis entirely.
@@ -200,9 +201,9 @@ QUERY TOOLS (each call costs 1 Pro Search query unless noted):
   All query tools accept source_focus: "none", "web", "academic", "social",
   "finance", "all". Use "none" for model-only queries without web search.
 
-  All query tools also accept an optional `conversation_id` (str) parameter. 
-  The server returns `[Conversation ID: <uuid>]` at the end of each response. 
-  Extract this UUID and pass it to the next query to maintain context across 
+  All query tools also accept an optional `conversation_id` (str) parameter.
+  The server returns `[Conversation ID: <uuid>]` at the end of each response.
+  Extract this UUID and pass it to the next query to maintain context across
   multiple turns. State is retained in memory for 1 hour.
 
 USAGE TOOL (1):
@@ -271,8 +272,8 @@ weekly pool (~300). Deep Research draws from a tiny monthly pool (~5-10).
 Wasting Pro queries on simple lookups means nothing left for real questions.
 
 COST MODEL:
-  Sonar 2 (pplx_sonar,     In-house model; still uses your session and Perplexity
-    quick intent)         counters — check pplx_usage() / pwm usage.
+  Sonar 2 (pplx_sonar,     In-house model; uses concise search mode to guarantee responses
+    quick intent)         are grounded. Decrements Perplexity session counters.
   Pro Search (standard,   Typically 1 from weekly Pro Search pool (~300/week
     detailed, pplx_ask,   on Pro/Max; exact rules are enforced by Perplexity).
     pplx_query, premium
@@ -283,11 +284,14 @@ COST MODEL:
 MANDATORY PROTOCOL:
   1. CHECK QUOTA FIRST: Call pplx_usage() before your first query each session.
   2. DEFAULT TO QUICK: Use pplx_smart_query(intent='quick') for most lookups.
-     It prefers Sonar 2 first and only escalates when the query needs a premium model.
+     It prefers Sonar 2 first (using concise mode to guarantee grounding) and
+     only escalates when the query needs a premium model.
   3. ESCALATE ONLY WHEN NEEDED: Use 'standard' for multi-source synthesis,
      'detailed' for complex analysis, 'research' only when user requests it.
   4. NEVER USE DEEP RESEARCH AUTONOMOUSLY — always ask the user first.
-  5. COUNCIL: Before calling pplx_council, ASK the user which models and how
+  5. SUBSCRIPTION-AWARE MODELS: Read the Subscription line from pplx_usage().
+     If it is Pro, exclude Max-only models: gpt55 and claude_opus.
+  6. COUNCIL: Before calling pplx_council, ASK the user which models and how
      many. Each model = 1 Pro Search. List available models for them to choose.
 
 WHEN TO USE EACH INTENT:
@@ -296,7 +300,7 @@ WHEN TO USE EACH INTENT:
   detailed  Complex analysis, deep reasoning, premium model       → 1 Pro
   research  Comprehensive reports (user must request explicitly)  → 1 Research
 
-DECISION RULE: Ask "Can Sonar 2 answer this?" If yes → quick. If no → standard.
+DECISION RULE: Ask "Can Sonar 2 answer this?" If yes → quick (grounded on concise search). If no → standard.
 Only use detailed/research when the complexity genuinely demands it.
 When in doubt, start with quick and escalate if the answer is insufficient.
 

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/cli/main.py

@@ -27,14 +27,16 @@ import rich_click as click
 
 from perplexity_web_mcp.exceptions import AuthenticationError, RateLimitError
 from perplexity_web_mcp.shared import (
+    COUNCIL_DEFAULT_MODELS_STR,
     COUNCIL_DISPLAY_NAMES,
+    COUNCIL_ELIGIBLE_MODEL_NAMES,
     MODEL_MAP,
     MODEL_NAMES,
     SOURCE_FOCUS_NAMES,
-    THINKING_TOGGLEABLE,
     Models,
     SourceFocusName,
     ask,
+    build_council_model_list,
     get_limit_cache,
     resolve_model,
 )
@@ -230,7 +232,7 @@ def _cmd_research_impl(query, source, json_output):
 
 # ── Council ────────────────────────────────────────────────────────────────
 
-COUNCIL_MODEL_NAMES = tuple(n for n in MODEL_NAMES if n not in {"auto", "sonar", "deep_research"})
+COUNCIL_MODEL_NAMES = COUNCIL_ELIGIBLE_MODEL_NAMES
 
 
 @cli.command()
@@ -239,7 +241,7 @@ COUNCIL_MODEL_NAMES = tuple(n for n in MODEL_NAMES if n not in {"auto", "sonar",
     "-m",
     "--models",
     "models_str",
-    default="gpt54,claude_opus,gemini_pro",
+    default=COUNCIL_DEFAULT_MODELS_STR,
     help=f"Comma-separated models ({', '.join(COUNCIL_MODEL_NAMES)}).",
 )
 @click.option("-t", "--thinking", is_flag=True, help="Enable extended thinking mode.")
@@ -301,14 +303,8 @@ def _cmd_council_impl(query, models_str, source, synthesize, json_output, thinki
 
         # Build model list (None = use defaults)
         model_list = None
-        if models_str != "gpt54,claude_opus,gemini_pro":
-            model_list = []
-            for name in model_names:
-                resolved = resolve_model(name, thinking=thinking)
-                display = COUNCIL_DISPLAY_NAMES.get(name, name)
-                if thinking and name in THINKING_TOGGLEABLE:
-                    display += " Thinking"
-                model_list.append((display, resolved))
+        if models_str != COUNCIL_DEFAULT_MODELS_STR:
+            model_list = build_council_model_list(model_names, thinking=thinking)
 
         synthesis_model = resolve_model(chairman) if chairman != "sonar" else None
 
@@ -454,16 +450,23 @@ def _cmd_usage_impl(refresh):
 
     # ── Account Info ───────────────────────────────────────────────────────
     settings = cache.get_user_settings(force_refresh=refresh)
-    if settings:
+    from perplexity_web_mcp.cli.auth import get_user_info
+
+    user_info = get_user_info(token)
+    if settings or user_info:
         table = Table(title="👤 Account", show_header=True, header_style="bold cyan")
         table.add_column("Field", style="bold")
         table.add_column("Value", justify="right")
 
-        tier = (settings.subscription_tier or "unknown").title()
-        status = settings.subscription_status
-        table.add_row("Subscription", f"[bold]{tier}[/] ({status})")
-        table.add_row("Total Queries", f"{settings.query_count:,}")
-        table.add_row("Pro Queries", f"{settings.query_count_copilot:,}")
+        if user_info:
+            table.add_row("Subscription", f"[bold]{user_info.tier_display}[/]")
+
+        if settings:
+            billing = settings.subscription_tier or "unknown"
+            status = settings.subscription_status
+            table.add_row("Billing", f"[bold]{billing}[/] ({status})")
+            table.add_row("Total Queries", f"{settings.query_count:,}")
+            table.add_row("Pro Queries", f"{settings.query_count_copilot:,}")
 
         console.print(table)
 
@@ -732,7 +735,7 @@ def _cmd_council(args: list[str]) -> int:
         return 1
 
     query = args[0]
-    models_str = "gpt54,claude_opus,gemini_pro"
+    models_str = COUNCIL_DEFAULT_MODELS_STR
     source: SourceFocusName = "web"
     synthesize = True
     json_output = False

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/council.py

@@ -16,6 +16,7 @@ from .config import ConversationConfig
 from .enums import CitationMode, SearchFocus, SourceFocus
 from .logging import get_logger
 from .models import Model, Models
+from .shared import COUNCIL_DEFAULT_MODEL_NAMES, build_council_model_list
 
 
 if TYPE_CHECKING:
@@ -29,19 +30,14 @@ logger = get_logger(__name__)
 # Default council composition
 # ---------------------------------------------------------------------------
 
-COUNCIL_DEFAULT_MODELS: list[tuple[str, Model]] = [
-    ("GPT-5.4", Models.GPT_54),
-    ("Claude Opus 4.7", Models.CLAUDE_47_OPUS),
-    ("Gemini 3.1 Pro", Models.GEMINI_31_PRO_THINKING),
-]
-"""Default models for the council (3 diverse providers)."""
+COUNCIL_DEFAULT_MODELS: list[tuple[str, Model]] = build_council_model_list(COUNCIL_DEFAULT_MODEL_NAMES)
+"""Default Pro-compatible models for the council (3 diverse providers)."""
 
-COUNCIL_DEFAULT_MODELS_THINKING: list[tuple[str, Model]] = [
-    ("GPT-5.4 Thinking", Models.GPT_54_THINKING),
-    ("Claude Opus 4.7 Thinking", Models.CLAUDE_47_OPUS_THINKING),
-    ("Gemini 3.1 Pro", Models.GEMINI_31_PRO_THINKING),
-]
-"""Default models for the council with extended thinking enabled."""
+COUNCIL_DEFAULT_MODELS_THINKING: list[tuple[str, Model]] = build_council_model_list(
+    COUNCIL_DEFAULT_MODEL_NAMES,
+    thinking=True,
+)
+"""Default Pro-compatible models for the council with extended thinking enabled."""
 
 
 # ---------------------------------------------------------------------------
@@ -220,7 +216,7 @@ def council_ask(
     Args:
         query: The question to ask all models.
         models: List of (display_name, Model) tuples. Defaults to
-                COUNCIL_DEFAULT_MODELS (GPT-5.4, Claude Opus, Gemini Pro).
+                COUNCIL_DEFAULT_MODELS (GPT-5.4, Claude Sonnet, Gemini Pro).
         source_focus: Source focus for all queries (none/web/academic/social/finance/all).
         synthesize: Whether to produce a synthesized consensus (adds 1 Sonar 2 synthesis query by default).
         thinking: Use thinking model variants for default council members.

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/data/SKILL.md

@@ -2,7 +2,7 @@
 name: perplexity-web-mcp
 description: "Search the web and query AI models via Perplexity AI using perplexity-web-mcp-cli. Supports CLI commands (pwm ask, pwm research), MCP tools (pplx_*), and Anthropic/OpenAI-compatible API server. Use when the user mentions \"perplexity\", \"pplx\", \"pwm\", \"web search with AI\", \"deep research\", \"search the internet\", or wants to query premium models like GPT-5.4, GPT-5.5, Claude, Gemini, Nemotron through Perplexity's web interface."
 metadata:
-  version: "0.10.7"
+  version: "0.12.2"
   author: "Jacob BD"
 ---
 
@@ -46,8 +46,9 @@ the weekly pool fast, leaving nothing for questions that actually need it.
 ### Before Every Session
 
 1. **Check quota first**: Call `pplx_usage()` (MCP) or `pwm usage` (CLI) before your first query.
-2. Review the remaining Pro and Research counts.
-3. If Pro < 20% remaining, restrict yourself to quick/Sonar 2 for everything except user-requested Pro queries.
+2. Review the remaining Pro and Research counts and the `Subscription` line.
+3. If Subscription is Pro, exclude Max-only models (`gpt55`, `claude_opus`) from model selection and councils.
+4. If Pro < 20% remaining, restrict yourself to quick/Sonar 2 for everything except user-requested Pro queries.
 
 ### Before Every Query: Choose the Lowest Sufficient Tier
 
@@ -83,8 +84,9 @@ Ask yourself: **"Can Sonar 2 answer this?"** If yes, use `quick`. Only escalate
 - The user needs high-confidence answers validated across multiple AI providers
 - Important decisions, fact-checking, or complex analysis
 - BEFORE calling: ASK the user which models and how many (each = 1 Pro Search)
-- Available models: gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26
-- Default: 3 models (GPT-5.4, Claude Opus, Gemini Pro) + synthesis = 4 Pro Searches
+- Available models: sonar, gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26
+- Max-only models: gpt55, claude_opus. Do not use these for Pro subscriptions.
+- Default: 3 Pro-compatible models (GPT-5.4, Claude Sonnet, Gemini Pro) + synthesis = 4 Pro Searches
 
 ### Decision Flowchart
 
@@ -127,7 +129,7 @@ The smart router automatically protects you:
 - **Healthy quota**: Uses the ideal model for your intent
 - **Low quota (<20% pro remaining)**: Response footer warns you to conserve
 - **Critical quota (<10% pro remaining)**: Downgrades detailed→auto to conserve
-- **Exhausted quota**: Falls back to Sonar 2 for everything except research
+- **Exhausted quota**: Falls back to Sonar 2 for everything except research (Sonar 2 is forced to concise mode to ensure grounded responses using search results)
 - **Research exhausted**: Falls back to premium Pro Search
 - Response metadata shows what model was used, why, and remaining quota
 
@@ -237,7 +239,8 @@ pwm ask "protein folding advances" -m gemini_pro -s academic --json
 ### Model Council
 
 Query multiple models in parallel and get a synthesized consensus.
-Each model in the council costs 1 Pro Search, plus 1 for Sonar 2 synthesis. Default: 3 models + synthesis = 4 Pro Searches.
+Each model in the council costs 1 Pro Search, plus 1 for Sonar 2 synthesis. Default: 3 Pro-compatible models + synthesis = 4 Pro Searches.
+Before selecting models, check `pplx_usage()` or `pwm usage`. If the subscription is Pro, exclude Max-only models (`gpt55`, `claude_opus`).
 
 ```bash
 pwm council "What are the best practices for microservices?"           # default 3 models
@@ -283,7 +286,7 @@ pwm usage --refresh         # Force-refresh from server
 | `pplx_sonar` | 1 Pro Search | Perplexity Sonar 2 |
 | `pplx_query` | 1 Pro | Explicit model selection with thinking toggle |
 | `pplx_ask` | 1 Pro | Quick Q&A (auto model) |
-| `pplx_council` | **N+1 Pro** (1 per model + 1 synthesis) | Model Council — **ASK USER which models first!** Supports `thinking=True` and `chairman` for synthesis model. |
+| `pplx_council` | **N+1 Pro** (1 per model + 1 synthesis) | Model Council — **ASK USER which models first!** Check subscription first; exclude Max-only `gpt55`/`claude_opus` on Pro. Supports `thinking=True` and `chairman` for synthesis model. |
 | `pplx_gpt54` / `_thinking` | 1 Pro | OpenAI GPT-5.4 (versatile) |
 | `pplx_gpt55` / `_thinking` | 1 Pro | OpenAI GPT-5.5 (latest, Max tier) |
 | `pplx_claude_sonnet` / `_think` | 1 Pro | Anthropic Claude 4.6 Sonnet |
@@ -309,7 +312,7 @@ For full MCP tool parameters: See [references/mcp-tools.md](references/mcp-tools
 | CLI Name | Provider | Thinking | Notes |
 |----------|----------|----------|-------|
 | auto | Perplexity | No | Auto-selects best |
-| sonar | Perplexity | No | Sonar 2 (API id `experimental`) |
+| sonar | Perplexity | No | Sonar 2 (API id `experimental`). Uses `mode="concise"` to ensure grounded answers. |
 | deep_research | Perplexity | No | Monthly quota |
 | gpt54 | OpenAI | Toggle | GPT-5.4 (versatile) |
 | gpt55 | OpenAI | Toggle | GPT-5.5 (latest, Max tier) |

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/data/references/mcp-tools.md

@@ -120,7 +120,7 @@ Returns a summary including:
 - Deep Research remaining (monthly)
 - Create Files & Apps remaining (monthly)
 - Browser Agent remaining (monthly)
-- Subscription tier and account info
+- Subscription tier, billing detail, and account info
 
 ## Authentication Tools
 

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/data/references/models.md

@@ -16,7 +16,7 @@ Complete list of models available through Perplexity Web MCP.
 - **Thinking:** No
 - **CLI:** `pwm ask "query" -m sonar`
 - **MCP:** `pplx_sonar(query)` or `pplx_query(query, model="sonar")`
-- **Notes:** Perplexity's latest in-house model.
+- **Notes:** Perplexity's latest in-house model. Settings default to concise search mode (`mode="concise"`), which bypasses interactive copilot to guarantee responses are grounded on retrieved search citations on all accounts (including Free tier fallback).
 
 ### deep_research (Deep Research)
 - **Identifier:** `pplx_alpha`

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/mcp/server.py

@@ -15,11 +15,11 @@ from fastmcp import FastMCP
 
 from perplexity_web_mcp.models import Models
 from perplexity_web_mcp.shared import (
-    COUNCIL_DISPLAY_NAMES,
-    THINKING_TOGGLEABLE,
+    COUNCIL_DEFAULT_MODELS_STR,
     ModelName,
     SourceFocusName,
     ask,
+    build_council_model_list,
     council_ask,
     get_limit_cache,
     resolve_model,
@@ -39,6 +39,7 @@ mcp = FastMCP(
         "- pplx_deep_research: 1 DEEP RESEARCH each (small monthly pool, ~5-10 total)\n\n"
         "MANDATORY PROTOCOL:\n"
         "1. On your FIRST query of the session, call pplx_usage() to check remaining quotas.\n"
+        "   Read the Subscription line: Pro users must avoid Max-only models.\n"
         "2. DEFAULT to pplx_smart_query(intent='quick') for most lookups — it prefers Sonar 2 "
         "before premium models when that fits the question.\n"
         "3. Only use 'standard' or 'detailed' intent when the question requires synthesis, "
@@ -288,7 +289,7 @@ def pplx_smart_query(
 def pplx_council(
     query: str,
     source_focus: SourceFocusName = "web",
-    models: str = "gpt54,claude_opus,gemini_pro",
+    models: str = COUNCIL_DEFAULT_MODELS_STR,
     synthesize: bool = True,
     thinking: bool = False,
     chairman: ModelName = "sonar",
@@ -296,20 +297,22 @@ def pplx_council(
     """Model Council — query multiple models in parallel, get synthesized consensus.
 
     IMPORTANT — BEFORE calling this tool, you MUST:
-    1. Tell the user the available models: gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26
-    2. Ask the user WHICH models they want in their council and HOW MANY
-    3. Inform them of the cost: each council model = 1 Pro Search query, plus synthesis
+    1. Tell the user the available models: sonar, gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26
+    2. Check pplx_usage() first. If Subscription is Pro, do not include Max-only models: gpt55, claude_opus
+    3. Ask the user WHICH models they want in their council and HOW MANY
+    4. Inform them of the cost: each council model = 1 Pro Search query, plus synthesis
        (default chairman sonar = Sonar 2 pass — still counts as a normal query toward limits)
-    4. Get explicit confirmation before executing
+    5. Get explicit confirmation before executing
 
-    Default council: GPT-5.4, Claude Opus 4.7, Gemini 3.1 Pro (3 diverse providers).
+    Default council: GPT-5.4, Claude Sonnet 4.6, Gemini 3.1 Pro (Pro-compatible, 3 diverse providers).
 
     Args:
         query: The question to ask all council models
         source_focus: Source type for all models (none/web/academic/social/finance/all)
         models: Comma-separated model names to use as council members.
-                Available: gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26.
-                Default: "gpt54,claude_opus,gemini_pro" (3 models + synthesis = 4 Pro Searches)
+                Available: sonar, gpt54, gpt55, claude_sonnet, claude_opus, gemini_pro, nemotron, kimi_k26.
+                Default: "gpt54,claude_sonnet,gemini_pro" (3 models + synthesis = 4 Pro Searches)
+                Max-only: gpt55, claude_opus. Exclude these when pplx_usage shows a Pro subscription.
         synthesize: Whether to synthesize a consensus from all responses.
                     Set false to get only individual responses (saves 1 Sonar 2 call).
         thinking: Enable extended thinking for council models (gpt54, gpt55, claude_sonnet,
@@ -319,15 +322,9 @@ def pplx_council(
     """
     # Parse custom model list if provided
     model_list = None
-    if models != "gpt54,claude_opus,gemini_pro":
-        model_list = []
-        for name in models.split(","):
-            name = name.strip()
-            resolved = resolve_model(name, thinking=thinking)
-            display = COUNCIL_DISPLAY_NAMES.get(name, name)
-            if thinking and name in THINKING_TOGGLEABLE:
-                display += " Thinking"
-            model_list.append((display, resolved))
+    if models != COUNCIL_DEFAULT_MODELS_STR:
+        model_names = [name.strip() for name in models.split(",") if name.strip()]
+        model_list = build_council_model_list(model_names, thinking=thinking)
 
     synthesis_model = resolve_model(chairman) if chairman != "sonar" else None
 
@@ -376,12 +373,26 @@ def pplx_usage(refresh: bool = False) -> str:
     else:
         parts.append("WARNING: Could not fetch rate limits (network error or token issue).")
 
+    from perplexity_web_mcp.cli.auth import get_user_info
+
+    user_info = get_user_info(token)
     settings = cache.get_user_settings(force_refresh=refresh)
-    if settings:
+    if settings or user_info:
         parts.append("")
         parts.append("ACCOUNT INFO")
         parts.append("=" * 40)
-        parts.append(settings.format_summary())
+        if user_info:
+            parts.append(f"Subscription: {user_info.tier_display}")
+        if settings:
+            parts.append(f"Billing: {settings.subscription_tier} ({settings.subscription_status})")
+            parts.append(f"Total queries: {settings.query_count:,}")
+            parts.append(f"Pro queries: {settings.query_count_copilot:,}")
+            parts.append(f"Upload limit: {settings.upload_limit} files")
+            parts.append(f"Create limit: {settings.create_limit}")
+            parts.append(f"Pages limit: {settings.pages_limit}")
+            parts.append(f"Max files/user: {settings.max_files_per_user:,}")
+            parts.append(f"Max file size: {settings.connector_limits.max_file_size_mb} MB")
+            parts.append(f"Daily attachments: {settings.connector_limits.daily_attachment_limit}")
 
     credits = cache.get_credits(force_refresh=refresh)
     if credits:

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/models.py

@@ -25,7 +25,7 @@ class Models:
     BEST = Model(identifier="pplx_pro")
     """Best - Automatically selects the best model based on the query."""
 
-    SONAR = Model(identifier="experimental")
+    SONAR = Model(identifier="experimental", mode="concise")
     """Sonar 2 — Perplexity's latest in-house model (backend id: experimental)."""
 
     GEMINI_31_PRO_THINKING = Model(identifier="gemini31pro_high")

{perplexity_web_mcp_cli-0.12.0 → perplexity_web_mcp_cli-0.12.2}/src/perplexity_web_mcp/shared.py

@@ -7,6 +7,7 @@ Both the MCP server (mcp/server.py) and CLI (cli/main.py) import from here.
 
 from __future__ import annotations
 
+from dataclasses import dataclass
 from threading import Lock
 from typing import TYPE_CHECKING, Literal
 from uuid import uuid4
@@ -22,6 +23,7 @@ from .token_store import get_token_or_raise, load_token
 
 
 if TYPE_CHECKING:
+    from .council import CouncilResponse
     from .types import SearchResultItem
 
 
@@ -29,6 +31,20 @@ if TYPE_CHECKING:
 # Model and source focus mappings (single source of truth)
 # ---------------------------------------------------------------------------
 
+SubscriptionMinimumTier = Literal["free", "pro", "max"]
+
+
+@dataclass(frozen=True, slots=True)
+class ModelDefinition:
+    """Metadata and model instances for one user-facing model key."""
+
+    base_model: Model
+    thinking_model: Model | None
+    display_name: str
+    provider: str
+    minimum_tier: SubscriptionMinimumTier = "pro"
+    council_eligible: bool = True
+
 SOURCE_FOCUS_MAP: dict[str, list[SourceFocus]] = {
     "none": [],
     "web": [SourceFocus.WEB],
@@ -38,18 +54,49 @@ SOURCE_FOCUS_MAP: dict[str, list[SourceFocus]] = {
     "all": [SourceFocus.WEB, SourceFocus.ACADEMIC, SourceFocus.SOCIAL],
 }
 
+MODEL_METADATA: dict[str, ModelDefinition] = {
+    "auto": ModelDefinition(Models.BEST, None, "Auto (Best)", "Perplexity", council_eligible=False),
+    "sonar": ModelDefinition(Models.SONAR, None, "Sonar 2", "Perplexity"),
+    "deep_research": ModelDefinition(
+        Models.DEEP_RESEARCH,
+        None,
+        "Deep Research",
+        "Perplexity",
+        council_eligible=False,
+    ),
+    "gpt54": ModelDefinition(Models.GPT_54, Models.GPT_54_THINKING, "GPT-5.4", "OpenAI"),
+    "gpt55": ModelDefinition(Models.GPT_55, Models.GPT_55_THINKING, "GPT-5.5", "OpenAI", minimum_tier="max"),
+    "claude_sonnet": ModelDefinition(
+        Models.CLAUDE_46_SONNET,
+        Models.CLAUDE_46_SONNET_THINKING,
+        "Claude Sonnet 4.6",
+        "Anthropic",
+    ),
+    "claude_opus": ModelDefinition(
+        Models.CLAUDE_47_OPUS,
+        Models.CLAUDE_47_OPUS_THINKING,
+        "Claude Opus 4.7",
+        "Anthropic",
+        minimum_tier="max",
+    ),
+    "gemini_pro": ModelDefinition(
+        Models.GEMINI_31_PRO_THINKING,
+        Models.GEMINI_31_PRO_THINKING,
+        "Gemini 3.1 Pro",
+        "Google",
+    ),
+    "nemotron": ModelDefinition(
+        Models.NEMOTRON_3_SUPER,
+        Models.NEMOTRON_3_SUPER,
+        "Nemotron 3 Super",
+        "NVIDIA",
+    ),
+    "kimi_k26": ModelDefinition(Models.KIMI_K2_6, Models.KIMI_K2_6_THINKING, "Kimi K2.6", "Moonshot"),
+}
+"""User-facing model metadata. Update this table when model names or tier availability changes."""
+
 MODEL_MAP: dict[str, tuple[Model, Model | None]] = {
-    # (base_model, thinking_model) - None if no thinking variant
-    "auto": (Models.BEST, None),
-    "sonar": (Models.SONAR, None),
-    "deep_research": (Models.DEEP_RESEARCH, None),
-    "gpt54": (Models.GPT_54, Models.GPT_54_THINKING),
-    "gpt55": (Models.GPT_55, Models.GPT_55_THINKING),
-    "claude_sonnet": (Models.CLAUDE_46_SONNET, Models.CLAUDE_46_SONNET_THINKING),
-    "claude_opus": (Models.CLAUDE_47_OPUS, Models.CLAUDE_47_OPUS_THINKING),
-    "gemini_pro": (Models.GEMINI_31_PRO_THINKING, Models.GEMINI_31_PRO_THINKING),
-    "nemotron": (Models.NEMOTRON_3_SUPER, Models.NEMOTRON_3_SUPER),
-    "kimi_k26": (Models.KIMI_K2_6, Models.KIMI_K2_6_THINKING),
+    name: (definition.base_model, definition.thinking_model) for name, definition in MODEL_METADATA.items()
 }
 
 SourceFocusName = Literal["none", "web", "academic", "social", "finance", "all"]
@@ -70,21 +117,39 @@ MODEL_NAMES: list[str] = list(MODEL_MAP.keys())
 SOURCE_FOCUS_NAMES: list[str] = list(SOURCE_FOCUS_MAP.keys())
 
 COUNCIL_DISPLAY_NAMES: dict[str, str] = {
-    "auto": "Auto (Best)",
-    "sonar": "Sonar 2",
-    "gpt54": "GPT-5.4",
-    "gpt55": "GPT-5.5",
-    "claude_sonnet": "Claude Sonnet 4.6",
-    "claude_opus": "Claude Opus 4.7",
-    "gemini_pro": "Gemini 3.1 Pro",
-    "nemotron": "Nemotron 3 Super",
-    "kimi_k26": "Kimi K2.6",
+    name: definition.display_name for name, definition in MODEL_METADATA.items()
 }
 
 THINKING_TOGGLEABLE: frozenset[str] = frozenset(
     name for name, (base, thinking) in MODEL_MAP.items() if thinking is not None and thinking is not base
 )
 
+MAX_ONLY_MODEL_NAMES: frozenset[str] = frozenset(
+    name for name, definition in MODEL_METADATA.items() if definition.minimum_tier == "max"
+)
+
+COUNCIL_ELIGIBLE_MODEL_NAMES: tuple[str, ...] = tuple(
+    name for name, definition in MODEL_METADATA.items() if definition.council_eligible
+)
+
+COUNCIL_DEFAULT_MODEL_NAMES: tuple[str, ...] = ("gpt54", "claude_sonnet", "gemini_pro")
+COUNCIL_DEFAULT_MODELS_STR = ",".join(COUNCIL_DEFAULT_MODEL_NAMES)
+
+
+def build_council_model_list(
+    model_names: tuple[str, ...] | list[str],
+    thinking: bool = False,
+) -> list[tuple[str, Model]]:
+    """Build display/model pairs for council execution from model metadata."""
+    model_list: list[tuple[str, Model]] = []
+    for name in model_names:
+        resolved = resolve_model(name, thinking=thinking)
+        display = COUNCIL_DISPLAY_NAMES.get(name, name)
+        if thinking and name in THINKING_TOGGLEABLE:
+            display += " Thinking"
+        model_list.append((display, resolved))
+    return model_list
+
 
 def resolve_model(name: str, thinking: bool = False) -> Model:
     """Resolve a model name string to a Model instance.