UAT: CRP `ContextBudget` base class has weaker validation than spec — allows `max_tokens=0` and `reserved_tokens == max_tokens` #4839

New issue

Open

opened 2026-04-08 20:06:36 +00:00 by HAL9000 · 0 comments

HAL9000 commented

2026-04-08 20:06:36 +00:00

Owner

Bug Report

Feature Area: ACMS — Context Request Protocol (CRP), ContextBudget model

Severity: Medium — the CRP base class has two validation gaps that allow invalid budget configurations; code using the CRP model directly (not the core model) will silently accept invalid inputs

What Was Tested

Code-level analysis comparing src/cleveragents/domain/models/acms/crp.py against the spec (docs/reference/acms.md) and the core model (src/cleveragents/domain/models/core/context_fragment.py).

Expected Behavior (from spec)

From docs/reference/acms.md:

max_tokens -- Maximum total tokens (default 4096, must be >= 1)

reserved_tokens -- Tokens reserved for system prompt (default 512, must be >= 0)

Validation: reserved_tokens must be strictly less than max_tokens.

Actual Behavior

src/cleveragents/domain/models/acms/crp.py ContextBudget:

class ContextBudget(BaseModel, frozen=True):
    max_tokens: int = Field(
        ...,
        ge=0,          # ❌ allows 0 — spec says must be >= 1
        description="Maximum total token budget (>= 0)",
    )
    reserved_tokens: int = Field(
        default=0,
        ge=0,
        description="Reserved tokens (<= `max_tokens`)",
    )

    @model_validator(mode="after")
    def _validate_reservation(self) -> ContextBudget:
        if self.reserved_tokens > self.max_tokens:   # ❌ allows equal — spec says strictly less than
            raise ValueError("reserved_tokens cannot exceed max_tokens")
        return self

src/cleveragents/domain/models/core/context_fragment.py ContextBudget (correct):

class ContextBudget(CRPContextBudget, frozen=True):
    max_tokens: int = Field(ge=1, default=4096)   # ✅ correct: ge=1
    reserved_tokens: int = Field(ge=0, default=512)

    @model_validator(mode="after")
    def _check_reserved_within_max(self) -> ContextBudget:
        if self.reserved_tokens >= self.max_tokens:  # ✅ correct: >= (strictly less than)
            raise ValueError(...)
        return self

Specific Gaps

Field	CRP base class	Spec requirement	Core model
`max_tokens` constraint	`ge=0` (allows 0)	`>= 1`	`ge=1` ✅
`reserved_tokens` validation	`> max_tokens` raises (allows equal)	strictly less than	`>= max_tokens` raises ✅
`max_tokens` default	no default (required)	4096	`default=4096` ✅

Impact

ContextBudget(max_tokens=0) is accepted by the CRP base class — available_tokens would be 0, causing division-by-zero in budget_used calculation in ACMSPipeline.assemble()
ContextBudget(max_tokens=100, reserved_tokens=100) is accepted by the CRP base class — available_tokens would be 0, causing the pipeline to produce empty context without error
Code using from cleveragents.domain.models.acms.crp import ContextBudget directly (not the core model) will have these weaker validations

Code Locations

src/cleveragents/domain/models/acms/crp.py — CRP base class (wrong)
src/cleveragents/domain/models/core/context_fragment.py — core model (correct)

Fix Required

Update src/cleveragents/domain/models/acms/crp.py ContextBudget:

class ContextBudget(BaseModel, frozen=True):
    max_tokens: int = Field(
        ...,
        ge=1,          # Fix: spec says must be >= 1
        description="Maximum total token budget (>= 1)",
    )
    reserved_tokens: int = Field(
        default=0,
        ge=0,
        description="Reserved tokens (strictly less than max_tokens)",
    )

    @model_validator(mode="after")
    def _validate_reservation(self) -> ContextBudget:
        if self.reserved_tokens >= self.max_tokens:  # Fix: spec says strictly less than
            raise ValueError(
                f"reserved_tokens ({self.reserved_tokens}) must be strictly less than "
                f"max_tokens ({self.max_tokens})"
            )
        return self

Automated by CleverAgents Bot
Supervisor: UAT Testing | Agent: uat-tester

## Bug Report **Feature Area:** ACMS — Context Request Protocol (CRP), `ContextBudget` model **Severity:** Medium — the CRP base class has two validation gaps that allow invalid budget configurations; code using the CRP model directly (not the core model) will silently accept invalid inputs ### What Was Tested Code-level analysis comparing `src/cleveragents/domain/models/acms/crp.py` against the spec (`docs/reference/acms.md`) and the core model (`src/cleveragents/domain/models/core/context_fragment.py`). ### Expected Behavior (from spec) From `docs/reference/acms.md`: > - `max_tokens` -- Maximum total tokens (default 4096, **must be >= 1**) > - `reserved_tokens` -- Tokens reserved for system prompt (default 512, must be >= 0) > - **Validation: `reserved_tokens` must be strictly less than `max_tokens`.** ### Actual Behavior **`src/cleveragents/domain/models/acms/crp.py` `ContextBudget`:** ```python class ContextBudget(BaseModel, frozen=True): max_tokens: int = Field( ..., ge=0, # ❌ allows 0 — spec says must be >= 1 description="Maximum total token budget (>= 0)", ) reserved_tokens: int = Field( default=0, ge=0, description="Reserved tokens (<= `max_tokens`)", ) @model_validator(mode="after") def _validate_reservation(self) -> ContextBudget: if self.reserved_tokens > self.max_tokens: # ❌ allows equal — spec says strictly less than raise ValueError("reserved_tokens cannot exceed max_tokens") return self ``` **`src/cleveragents/domain/models/core/context_fragment.py` `ContextBudget` (correct):** ```python class ContextBudget(CRPContextBudget, frozen=True): max_tokens: int = Field(ge=1, default=4096) # ✅ correct: ge=1 reserved_tokens: int = Field(ge=0, default=512) @model_validator(mode="after") def _check_reserved_within_max(self) -> ContextBudget: if self.reserved_tokens >= self.max_tokens: # ✅ correct: >= (strictly less than) raise ValueError(...) return self ``` ### Specific Gaps | Field | CRP base class | Spec requirement | Core model | |---|---|---|---| | `max_tokens` constraint | `ge=0` (allows 0) | `>= 1` | `ge=1` ✅ | | `reserved_tokens` validation | `> max_tokens` raises (allows equal) | strictly less than | `>= max_tokens` raises ✅ | | `max_tokens` default | no default (required) | 4096 | `default=4096` ✅ | ### Impact - `ContextBudget(max_tokens=0)` is accepted by the CRP base class — `available_tokens` would be 0, causing division-by-zero in `budget_used` calculation in `ACMSPipeline.assemble()` - `ContextBudget(max_tokens=100, reserved_tokens=100)` is accepted by the CRP base class — `available_tokens` would be 0, causing the pipeline to produce empty context without error - Code using `from cleveragents.domain.models.acms.crp import ContextBudget` directly (not the core model) will have these weaker validations ### Code Locations - `src/cleveragents/domain/models/acms/crp.py` — CRP base class (wrong) - `src/cleveragents/domain/models/core/context_fragment.py` — core model (correct) ### Fix Required Update `src/cleveragents/domain/models/acms/crp.py` `ContextBudget`: ```python class ContextBudget(BaseModel, frozen=True): max_tokens: int = Field( ..., ge=1, # Fix: spec says must be >= 1 description="Maximum total token budget (>= 1)", ) reserved_tokens: int = Field( default=0, ge=0, description="Reserved tokens (strictly less than max_tokens)", ) @model_validator(mode="after") def _validate_reservation(self) -> ContextBudget: if self.reserved_tokens >= self.max_tokens: # Fix: spec says strictly less than raise ValueError( f"reserved_tokens ({self.reserved_tokens}) must be strictly less than " f"max_tokens ({self.max_tokens})" ) return self ``` --- **Automated by CleverAgents Bot** Supervisor: UAT Testing | Agent: uat-tester

HAL9000 referenced this issue

2026-04-08 20:08:39 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue

2026-04-08 20:16:04 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 added this to the v3.5.0 milestone

2026-04-08 20:18:42 +00:00

HAL9000 referenced this issue

2026-04-08 20:19:20 +00:00

[Automated] CleverAgents Build Session - 2026-04-08 #4799

HAL9000 referenced this issue