tokenjam 0.3.2__tar.gz → 0.3.3__tar.gz

{tokenjam-0.3.2 → tokenjam-0.3.3}/CLAUDE.md

@@ -191,9 +191,9 @@ The Agent Incident Library at `incidents/` is separate: each scenario is a `scen
 
 ## Pricing
 
-Model pricing lives in `pricing/models.toml` (USD per million tokens). Structure: `[provider.model_name]` with `input_per_mtok`, `output_per_mtok`, and optional `cache_read_per_mtok`/`cache_write_per_mtok`. Unknown models fall back to default rates ($0.50/$2.00 per MTok) with a logged warning. The pricing table is LRU-cached at process startup — restart to pick up changes.
+Model pricing lives in `tokenjam/pricing/models.toml` (USD per million tokens) — the packaged file `core/pricing.py` loads via `PRICING_FILE = Path(__file__).parent.parent / "pricing" / "models.toml"`. There is no repo-root `pricing/` copy (it was moved into the package in v0.1.x so it ships in the wheel; editing a repo-root file would have no runtime effect). Structure: `[provider.model_name]` with `input_per_mtok`, `output_per_mtok`, and optional `cache_read_per_mtok`/`cache_write_per_mtok`. Unknown models fall back to default rates ($0.50/$2.00 per MTok) with a logged warning. The pricing table is LRU-cached at process startup — restart to pick up changes.
 
-Pricing is community-maintained: submit a PR editing `pricing/models.toml` when provider prices change. No code changes needed — the file is loaded at runtime.
+Pricing is community-maintained: submit a PR editing `tokenjam/pricing/models.toml` when provider prices change. No code changes needed — the file is loaded at runtime.
 
 ## CI
 

{tokenjam-0.3.2 → tokenjam-0.3.3}/CONTRIBUTING.md

@@ -43,7 +43,7 @@ tokenjam/sdk/               @watch() decorator and provider/framework patches
 tokenjam/otel/              OTel TracerProvider and span exporter wiring
 tokenjam/utils/             Formatting, time parsing, ID generation
 sdk-ts/src/            TypeScript SDK (@tokenjam/sdk)
-pricing/models.toml    Community-maintained model pricing — PRs welcome here
+tokenjam/pricing/models.toml  Community-maintained model pricing — PRs welcome here
 tests/factories.py     Span factory — use this in all synthetic tests, never
                        construct NormalizedSpan directly
 ```
@@ -57,7 +57,7 @@ This project was built using parallel Claude Code agents. The `.claude/` directo
 
 ## Pricing table contributions
 
-The file `pricing/models.toml` is intentionally community-maintained. If a model is missing or prices have changed, open a PR with the update — no issue needed, just update the TOML and verify the format matches existing entries.
+The file `tokenjam/pricing/models.toml` is intentionally community-maintained. If a model is missing or prices have changed, open a PR with the update — no issue needed, just update the TOML and verify the format matches existing entries. (This is the file the cost engine loads at runtime; there is no separate repo-root copy.)
 
 ## Reporting issues
 

{tokenjam-0.3.2 → tokenjam-0.3.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tokenjam
-Version: 0.3.2
+Version: 0.3.3
 Summary: TokenJam — local-first OTel-native observability for Autonomous AI agents
 Project-URL: Homepage, https://opencla.watch
 Project-URL: Repository, https://github.com/Metabuilder-Labs/openclawwatch

{tokenjam-0.3.2 → tokenjam-0.3.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "tokenjam"
-version = "0.3.2"
+version = "0.3.3"
 description = "TokenJam — local-first OTel-native observability for Autonomous AI agents"
 readme = "README.md"
 requires-python = ">=3.10"

{tokenjam-0.3.2 → tokenjam-0.3.3}/sdk-ts/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tokenjam/sdk",
-  "version": "0.3.2",
+  "version": "0.3.3",
   "description": "TypeScript SDK for TokenJam — local-first observability for AI agents",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",

{tokenjam-0.3.2 → tokenjam-0.3.3}/tests/unit/test_cmd_tokenmaxx.py

@@ -14,27 +14,54 @@ from tokenjam.core.config import ProviderBudget, TjConfig
 # ───────────────────────────── classification ─────────────────────────────
 
 def test_classify_zero_spend_is_sipper():
-    t = _classify(0.0)
-    assert t.label == "TokenSipper"
+    # Zero spend → Sipper, regardless of which path.
+    assert _classify(0.0).label == "TokenSipper"
+    assert _classify(0.0, multiplier=0.0).label == "TokenSipper"
 
 
-def test_classify_walks_tiers_correctly():
-    # Each threshold should map into the next tier exactly at the boundary.
+def test_classify_multiplier_path_walks_tier_boundaries():
+    # Multiplier-based classification — the primary path for subscription users.
+    # Boundaries are 1× / 4× / 10× / 20× per the launch-tier spec.
     cases = [
-        (49.99,   "TokenSipper"),
-        (50.0,    "TokenModerator"),
-        (199.99,  "TokenModerator"),
-        (200.0,   "TokenMaxxer"),
-        (499.99,  "TokenMaxxer"),
-        (500.0,   "TokenChad"),
-        (1499.99, "TokenChad"),
-        (1500.0,  "TokenGigaChad"),
+        (0.99,  "TokenSipper"),
+        (1.0,   "TokenModerator"),
+        (3.99,  "TokenModerator"),
+        (4.0,   "TokenMaxxer"),
+        (9.99,  "TokenMaxxer"),
+        (10.0,  "TokenChad"),
+        (19.99, "TokenChad"),
+        (20.0,  "TokenGigaChad"),
+        (200.0, "TokenGigaChad"),
+    ]
+    for mult, expected in cases:
+        assert _classify(0.0, multiplier=mult).label == expected, f"failed at {mult}×"
+
+
+def test_classify_absolute_path_for_api_users_walks_tier_boundaries():
+    # Absolute USD/mo fallback — calibrated against Max-5x = $100/mo so the
+    # tier names mean roughly the same thing in both paths.
+    cases = [
+        (99.99,   "TokenSipper"),
+        (100.0,   "TokenModerator"),
+        (399.99,  "TokenModerator"),
+        (400.0,   "TokenMaxxer"),
+        (999.99,  "TokenMaxxer"),
+        (1000.0,  "TokenChad"),
+        (1999.99, "TokenChad"),
+        (2000.0,  "TokenGigaChad"),
         (50_000,  "TokenGigaChad"),
     ]
     for spend, expected in cases:
         assert _classify(spend).label == expected, f"failed at ${spend}"
 
 
+def test_classify_multiplier_overrides_absolute_when_both_provided():
+    # A subscription user at $50/mo on a $20 Pro plan (2.5× their plan) is
+    # a TokenModerator, NOT a TokenSipper — the multiplier path wins.
+    t = _classify(50.0, multiplier=2.5)
+    assert t.label == "TokenModerator"
+
+
 def test_every_tier_has_emoji_and_quip():
     # The artifact's social-shareability depends on every tier having
     # readable text — guard against a future blank entry.
@@ -95,3 +122,19 @@ def test_plan_label_and_fee_returns_none_for_unknown_or_api():
     assert _plan_label_and_fee("local") is None
     assert _plan_label_and_fee(None) is None
     assert _plan_label_and_fee("not_a_plan") is None
+
+
+def test_all_anthropic_subscription_plans_produce_a_multiplier():
+    # Every Anthropic subscription plan in the onboard wizard must produce
+    # a finite multiplier, otherwise the tokenmaxx tweet hook ("Nx your
+    # plan") doesn't render for those users. This guards against the table
+    # diverging from `cmd_onboard.py::_ANTHROPIC_PLAN_CHOICES`.
+    for plan in ("pro", "max_5x", "max_20x"):
+        info = _plan_label_and_fee(plan)
+        assert info is not None, f"{plan!r} not in plan-fee table"
+        label, fee = info
+        assert label, f"{plan!r} missing a display label"
+        assert fee and fee > 0, f"{plan!r} missing a numeric monthly fee"
+        # Sanity-check the multiplier math the renderer does.
+        spend = 1000.0
+        assert spend / fee > 0

{tokenjam-0.3.2 → tokenjam-0.3.3}/tests/unit/test_cost.py

@@ -113,6 +113,21 @@ def test_get_rates_opus_4_8():
     assert rates.cache_write_per_mtok == 6.25
 
 
+def test_calculate_cost_opus_4_5_model():
+    # claude-opus-4-5: input=5.00, output=25.00 (same tier as 4.6/4.7/4.8)
+    cost = calculate_cost("anthropic", "claude-opus-4-5", 1_000_000, 1_000_000)
+    assert cost == 30.0
+
+
+def test_get_rates_opus_4_5():
+    rates = get_rates("anthropic", "claude-opus-4-5")
+    assert rates is not None
+    assert rates.input_per_mtok == 5.00
+    assert rates.output_per_mtok == 25.00
+    assert rates.cache_read_per_mtok == 0.50
+    assert rates.cache_write_per_mtok == 6.25
+
+
 def test_calculate_cost_openai_model():
     # gpt-4o: input=2.50, output=10.00
     cost = calculate_cost("openai", "gpt-4o", 500_000, 100_000)

{tokenjam-0.3.2 → tokenjam-0.3.3}/tokenjam/cli/cmd_tokenmaxx.py

@@ -20,13 +20,17 @@ from dataclasses import dataclass
 
 import click
 
-from tokenjam.utils.formatting import console, format_cost
+from tokenjam.utils.formatting import console, format_cost  # noqa: F401  (kept for back-compat imports)
 from tokenjam.utils.time_parse import parse_since, utcnow
 
 
-# Tier table — monthly USD spend thresholds → (label, one-liner).
-# Calibrated for individual coding-agent users. Order matters: walked
-# bottom-up, first matching threshold wins.
+# Tier table — multiplier thresholds (×plan-fee) → (label, one-liner).
+# The `threshold` field is interpreted as a multiplier when the user has a
+# subscription plan declared, or as an absolute USD/mo amount when they
+# don't (API users, no plan set). The fallback thresholds in _SPEND_TIERS
+# below mirror the multiplier ladder calibrated against the Max-5x plan
+# ($100/mo) so the tier names mean roughly the same thing across paths.
+# Order matters: classify walks high-to-low; first match wins.
 @dataclass(frozen=True)
 class Tier:
     threshold: float
@@ -36,18 +40,34 @@ class Tier:
 
 
 _TIERS: list[Tier] = [
-    Tier(0,    "TokenSipper",     "💧", "Are you even using AI?"),
-    Tier(50,   "TokenModerator",  "🥱", "Mostly reasonable. Try harder."),
-    Tier(200,  "TokenMaxxer",     "💸", "You're paying Anthropic's rent."),
-    Tier(500,  "TokenChad",       "🔥", "You're paying their interns' rent too."),
-    Tier(1500, "TokenGigaChad",   "🔥🔥", "Touch grass. Then run `tj optimize`."),
+    Tier(0,  "TokenSipper",     "💧",   "Are you even using AI?"),
+    Tier(1,  "TokenModerator",  "🥱",   "Mostly reasonable. Try harder."),
+    Tier(4,  "TokenMaxxer",     "💸",   "You're paying Anthropic's rent."),
+    Tier(10, "TokenChad",       "🔥",   "You're paying their interns' rent too."),
+    Tier(20, "TokenGigaChad",   "🔥🔥", "Touch grass. Then run `tj optimize`."),
 ]
 
+# Absolute USD/mo fallback for users without a subscription plan (API users).
+# Calibrated against Max-5x = $100/mo: each threshold is the multiplier × $100.
+# This way a $400/mo API user and a 4× Pro/Max-5x/Max-20x user both end up
+# in TokenMaxxer — the tier name reflects "shocking spend" in either world.
+_SPEND_TIER_THRESHOLDS_USD: list[float] = [0, 100, 400, 1000, 2000]
 
-def _classify(monthly_spend: float) -> Tier:
-    """Walk tiers high-to-low; first threshold the spend exceeds wins."""
-    for tier in reversed(_TIERS):
-        if monthly_spend >= tier.threshold:
+
+def _classify(monthly_spend: float, multiplier: float | None = None) -> Tier:
+    """
+    Pick a tier. Prefer the multiplier path when the user has a subscription
+    plan with a declared fee; fall back to absolute monthly spend when not.
+    Walks tiers high-to-low; first matching threshold wins.
+    """
+    if multiplier is not None:
+        for tier in reversed(_TIERS):
+            if multiplier >= tier.threshold:
+                return tier
+        return _TIERS[0]
+    # API / no-plan path — map onto the same tier labels via absolute USD.
+    for tier, threshold_usd in zip(reversed(_TIERS), reversed(_SPEND_TIER_THRESHOLDS_USD)):
+        if monthly_spend >= threshold_usd:
             return tier
     return _TIERS[0]
 
@@ -86,7 +106,7 @@ def cmd_tokenmaxx(ctx: click.Context, since: str, output_json: bool) -> None:
     if plan_info and plan_info[1]:
         multiplier = monthly_spend / plan_info[1]
 
-    tier = _classify(monthly_spend)
+    tier = _classify(monthly_spend, multiplier)
 
     if output_json:
         click.echo(json.dumps({
@@ -201,6 +221,28 @@ def _plan_label_and_fee(plan_tier: str | None) -> tuple[str, float | None] | Non
 
 # ───────────────────────────── rendering ──────────────────────────────────
 
+def _fmt_spend(usd: float) -> str:
+    """Spend / savings: always 2 decimals — readable, screenshot-friendly.
+
+    The default `format_cost` helper uses 4 decimals (precision for the cost
+    engine internals); for the tokenmaxx social artifact we want $4044.57,
+    not $4044.5774.
+    """
+    return f"${usd:.2f}"
+
+
+def _fmt_fee(usd: float) -> str:
+    """Plan fee: drop the decimals when the fee is a round dollar.
+
+    Anthropic / OpenAI subscription tiers (Pro $20, Max-5x $100, Max-20x $200)
+    are all whole-dollar — `$100.00` reads worse than `$100`. Falls back to
+    2 decimals for anything fractional.
+    """
+    if usd == int(usd):
+        return f"${int(usd)}"
+    return f"${usd:.2f}"
+
+
 def _render(
     *, tier: Tier, spend_usd: float, monthly_spend: float,
     window_days: int, sessions: int,
@@ -208,7 +250,12 @@ def _render(
     multiplier: float | None,
     savings_usd: float,
 ) -> None:
-    """Big-headline render. Designed to be a clean screenshot artifact."""
+    """
+    Big-headline render. Designed to be a clean screenshot artifact:
+    bordered Panel with a heading, the tier callout up top, the spend
+    breakdown in the middle, the action line at the bottom, and the
+    share prompt OUTSIDE the panel.
+    """
     if sessions == 0:
         console.print(
             "\n[yellow]No usage data found.[/yellow]\n"
@@ -217,54 +264,96 @@ def _render(
         )
         return
 
-    # Banner — the headline shareable line.
-    console.print()
-    console.print(f"  {tier.emoji} [bold]You're a {tier.label}.[/bold]")
-    console.print(f"     [dim italic]\"{tier.quip}\"[/dim italic]")
-    console.print()
-
-    # Spend breakdown — what produced the tier.
+    # Build the inside-the-panel content as one rich Text/markup string.
+    # Rich Panel doesn't have native sub-spacing primitives, so we hand-pad
+    # with newlines to get the visual rhythm we want in the screenshot.
+    from rich.console import Group
+    from rich.panel import Panel
+    from rich.text import Text
+    from rich.align import Align
+
+    # Tier callout — the headline, plus a larger non-dim quip (no quotes)
+    # with `tj optimize` highlighted green-bold when it appears in the quip.
+    headline = Text()
+    headline.append(f"{tier.emoji} ", style="")
+    headline.append("You're a ", style="bold")
+    headline.append(tier.label, style="bold")
+    headline.append(".", style="bold")
+
+    quip_text = Text()
+    # Walk the quip and recolor any `tj optimize` backtick-wrapped token green.
+    # Rich doesn't auto-parse backticks, so we do it manually with split.
+    parts = tier.quip.split("`tj optimize`")
+    for i, p in enumerate(parts):
+        if p:
+            quip_text.append(p, style="")
+        if i < len(parts) - 1:
+            quip_text.append("tj optimize", style="bold green")
+
+    # The spend / multiplier block — same content as before but with the
+    # cleaner formatters and a slightly tighter line structure.
+    body = Text()
     actual_label = f"last {window_days}d" if window_days < 30 else "last 30d"
-    if window_days == 30:
-        console.print(
-            f"  [bold]{format_cost(spend_usd)}[/bold] in {actual_label} "
-            f"across [bold]{sessions}[/bold] sessions."
-        )
-    else:
-        console.print(
-            f"  [bold]{format_cost(spend_usd)}[/bold] in {actual_label} "
-            f"across [bold]{sessions}[/bold] sessions "
-            f"(≈ [bold]{format_cost(monthly_spend)}/mo[/bold] at this rate)."
-        )
+    body.append(_fmt_spend(spend_usd), style="bold")
+    body.append(f" in {actual_label} across ")
+    body.append(str(sessions), style="bold")
+    body.append(" sessions.")
+    if window_days != 30:
+        body.append("  (≈ ", style="dim")
+        body.append(_fmt_spend(monthly_spend), style="bold")
+        body.append("/mo at this rate)", style="dim")
 
-    # Plan multiplier — the punchline.
     if plan_info and multiplier:
         plan_label, fee = plan_info
-        console.print(
-            f"  That's [bold]{multiplier:.1f}×[/bold] your "
-            f"[bold]{plan_label}[/bold] cost ({format_cost(fee)}/mo flat)."
-        )
+        body.append("\nThat's ")
+        body.append(f"{multiplier:.1f}×", style="bold")
+        body.append(" your ")
+        body.append(plan_label, style="bold")
+        body.append(f" cost ({_fmt_fee(fee)}/mo flat).")
     elif plan_info:
         plan_label, _ = plan_info
-        console.print(f"  Plan: [bold]{plan_label}[/bold].")
-
-    console.print()
-
-    # The action — the part that makes this a tool, not a flex.
+        body.append("\nPlan: ")
+        body.append(plan_label, style="bold")
+        body.append(".")
+
+    # Action line — savings recoverable, or fall through to "no obvious
+    # savings yet". `tj optimize` rendered green-bold either way so the
+    # eye lands on the verb.
+    action = Text("💡 ")
     if savings_usd > 0:
-        console.print(
-            f"  💡 [bold]{format_cost(savings_usd)}/mo[/bold] of that looks "
-            f"recoverable. Run [bold]tj optimize[/bold] to see candidates."
-        )
+        action.append(_fmt_spend(savings_usd) + "/mo", style="bold")
+        action.append(" of that looks recoverable. Run ")
+        action.append("tj optimize", style="bold green")
+        action.append(" to see candidates.")
     else:
-        console.print(
-            "  💡 No obvious savings flagged yet — run [bold]tj optimize[/bold] "
-            "for the full report once you have more data."
-        )
-    console.print()
+        action.append("No obvious savings flagged yet — run ")
+        action.append("tj optimize", style="bold green")
+        action.append(" for the full report once you have more data.")
+
+    # Compose with deliberate vertical spacing.
+    panel_body = Group(
+        headline,
+        Text(""),            # blank line under headline
+        Align.left(quip_text),
+        Text(""),            # blank line under quip
+        body,
+        Text(""),            # blank line before action
+        action,
+    )
 
-    # Subtle share prompt.
+    console.print()
+    console.print(Panel(
+        panel_body,
+        title="[bold]TokenJam TokenMaxxing Report[/bold]",
+        title_align="left",
+        border_style="dim",
+        padding=(1, 2),
+    ))
+
+    # Share prompt — outside the panel, teal, points at the brand handle so
+    # the social mechanic routes to a real account we can amplify from.
     console.print(
-        "  [dim]Share your tier: screenshot the above and post with #tokenmaxx[/dim]"
+        "  [cyan]Share your tier: screenshot the above and tag "
+        "[bold]@tokenjamdev[/bold][/cyan]"
     )
     console.print()

{tokenjam-0.3.2 → tokenjam-0.3.3}/tokenjam/core/cost.py

@@ -32,7 +32,7 @@ def calculate_cost(
     if rates is None:
         logger.warning(
             "No pricing data for %s/%s — using default rates. "
-            "Add to pricing/models.toml to get accurate costs.",
+            "Add to tokenjam/pricing/models.toml to get accurate costs.",
             provider, model,
         )
         rates = ModelRates(

{tokenjam-0.3.2 → tokenjam-0.3.3}/tokenjam/pricing/models.toml

@@ -1,4 +1,4 @@
-# pricing/models.toml
+# tokenjam/pricing/models.toml
 # Prices in USD per million tokens.
 # Submit a PR when provider prices change.
 
@@ -15,10 +15,10 @@ cache_read_per_mtok = 0.50
 cache_write_per_mtok = 6.25
 
 [anthropic.claude-opus-4-5]
-input_per_mtok = 15.00
-output_per_mtok = 75.00
-cache_read_per_mtok = 1.50
-cache_write_per_mtok = 18.75
+input_per_mtok = 5.00
+output_per_mtok = 25.00
+cache_read_per_mtok = 0.50
+cache_write_per_mtok = 6.25
 
 [anthropic.claude-opus-4-6]
 input_per_mtok = 5.00

tokenjam-0.3.2/pricing/models.toml

@@ -1,82 +0,0 @@
-# pricing/models.toml
-# Prices in USD per million tokens.
-# Submit a PR when provider prices change.
-
-[anthropic.claude-opus-4-8]
-input_per_mtok = 5.00
-output_per_mtok = 25.00
-cache_read_per_mtok = 0.50
-cache_write_per_mtok = 6.25
-
-[anthropic.claude-opus-4-7]
-input_per_mtok = 5.00
-output_per_mtok = 25.00
-cache_read_per_mtok = 0.50
-cache_write_per_mtok = 6.25
-
-[anthropic.claude-opus-4-6]
-input_per_mtok = 5.00
-output_per_mtok = 25.00
-cache_read_per_mtok = 0.50
-cache_write_per_mtok = 6.25
-
-[anthropic.claude-sonnet-4-6]
-input_per_mtok = 3.00
-output_per_mtok = 15.00
-cache_read_per_mtok = 0.30
-cache_write_per_mtok = 3.75
-
-[anthropic.claude-haiku-4-5]
-input_per_mtok = 0.80
-output_per_mtok = 4.00
-cache_read_per_mtok = 0.08
-cache_write_per_mtok = 1.00
-
-[openai.gpt-4o]
-input_per_mtok = 2.50
-output_per_mtok = 10.00
-
-[openai.gpt-4o-mini]
-input_per_mtok = 0.15
-output_per_mtok = 0.60
-
-[openai.o3]
-input_per_mtok = 10.00
-output_per_mtok = 40.00
-
-[openai.o4-mini]
-input_per_mtok = 1.10
-output_per_mtok = 4.40
-
-[google.gemini-2-5-pro]
-input_per_mtok = 1.25
-output_per_mtok = 10.00
-
-[google.gemini-2-5-flash]
-input_per_mtok = 0.15
-output_per_mtok = 0.60
-
-[aws.us-amazon-nova-pro-v1]
-input_per_mtok = 0.80
-output_per_mtok = 3.20
-
-[aws.us-amazon-nova-lite-v1]
-input_per_mtok = 0.06
-output_per_mtok = 0.24
-
-# OpenAI-compatible providers (Groq, Together, Fireworks, xAI, Azure OpenAI)
-# use patch_openai() with a custom base_url — add their model names here
-# as they are encountered.
-
-[groq.llama-3-3-70b-versatile]
-input_per_mtok = 0.59
-output_per_mtok = 0.79
-
-[xai.grok-3]
-input_per_mtok = 3.00
-output_per_mtok = 15.00
-
-# HUD managed inference (inference.hud.ai)
-[hud.claude-sonnet-4-5]
-input_per_mtok = 3.00
-output_per_mtok = 15.00