open-pharma-plugins 2.2.0__py3-none-any.whl

open_pharma_plugins_field_training/fixtures/sample_product_message.pdf

@@ -0,0 +1,156 @@
+%PDF-1.4
+%���� ReportLab Generated PDF document (opensource)
+1 0 obj
+<<
+/F1 2 0 R /F2 3 0 R /F3 6 0 R
+>>
+endobj
+2 0 obj
+<<
+/BaseFont /Helvetica /Encoding /WinAnsiEncoding /Name /F1 /Subtype /Type1 /Type /Font
+>>
+endobj
+3 0 obj
+<<
+/BaseFont /Helvetica-Bold /Encoding /WinAnsiEncoding /Name /F2 /Subtype /Type1 /Type /Font
+>>
+endobj
+4 0 obj
+<<
+/Contents 13 0 R /MediaBox [ 0 0 612 792 ] /Parent 12 0 R /Resources <<
+/Font 1 0 R /ProcSet [ /PDF /Text /ImageB /ImageC /ImageI ]
+>> /Rotate 0 /Trans <<
+
+>> 
+  /Type /Page
+>>
+endobj
+5 0 obj
+<<
+/Contents 14 0 R /MediaBox [ 0 0 612 792 ] /Parent 12 0 R /Resources <<
+/Font 1 0 R /ProcSet [ /PDF /Text /ImageB /ImageC /ImageI ]
+>> /Rotate 0 /Trans <<
+
+>> 
+  /Type /Page
+>>
+endobj
+6 0 obj
+<<
+/BaseFont /Symbol /Name /F3 /Subtype /Type1 /Type /Font
+>>
+endobj
+7 0 obj
+<<
+/Contents 15 0 R /MediaBox [ 0 0 612 792 ] /Parent 12 0 R /Resources <<
+/Font 1 0 R /ProcSet [ /PDF /Text /ImageB /ImageC /ImageI ]
+>> /Rotate 0 /Trans <<
+
+>> 
+  /Type /Page
+>>
+endobj
+8 0 obj
+<<
+/Contents 16 0 R /MediaBox [ 0 0 612 792 ] /Parent 12 0 R /Resources <<
+/Font 1 0 R /ProcSet [ /PDF /Text /ImageB /ImageC /ImageI ]
+>> /Rotate 0 /Trans <<
+
+>> 
+  /Type /Page
+>>
+endobj
+9 0 obj
+<<
+/Contents 17 0 R /MediaBox [ 0 0 612 792 ] /Parent 12 0 R /Resources <<
+/Font 1 0 R /ProcSet [ /PDF /Text /ImageB /ImageC /ImageI ]
+>> /Rotate 0 /Trans <<
+
+>> 
+  /Type /Page
+>>
+endobj
+10 0 obj
+<<
+/PageMode /UseNone /Pages 12 0 R /Type /Catalog
+>>
+endobj
+11 0 obj
+<<
+/Author (\(anonymous\)) /CreationDate (D:20260819134834+08'00') /Creator (\(unspecified\)) /Keywords () /ModDate (D:20260819134834+08'00') /Producer (ReportLab PDF Library - \(opensource\)) 
+  /Subject (\(unspecified\)) /Title (\(anonymous\)) /Trapped /False
+>>
+endobj
+12 0 obj
+<<
+/Count 5 /Kids [ 4 0 R 5 0 R 7 0 R 8 0 R 9 0 R ] /Type /Pages
+>>
+endobj
+13 0 obj
+<<
+/Filter [ /ASCII85Decode /FlateDecode ] /Length 675
+>>
+stream
+Gatn"9on!^&;KZOMYG1m,.3VrKY>b:,V#,9X`EWVMb+>e8o-D/W0$l9CE1(2L.[`"%m(Y)4O6)FTcj')DTk`298_%/$L/jAJVIKeFG=ehH10+TB:L(gO`g*)rFtXHiPGBInL1)`Lr/b9^-;n^K`MYB!7eYj\#3:>e'^_WiO`cYE<]#;00ZJ73P$m8VjD>NmoXZ^Yu=(sZK:*)!U1,jT?t2n[>8`CJGS=k!6(':P/.`D\D:2#a)emo.^(X\B<F*rF+&eSG&N)/etF[B,kOk-$f$<fq$ui'=q3J/G_:Hc`1<h6[H:TZYnF^j;=[t&X8nGE10'XQb(P1#S<Y!b(AkFul[@K/KLqYj0_QFp+$nr*$ganJYX,sqhcGd)87cqcV879GLuIZ@cR41k,D6+.[em+nJFRBn1F%)@&RUs-=gc;dpJRG1JlYGDBp1saZ&ANY(X^<h,roYPFO!i,8mm=8-a5Lp^tp[nEOOU=iOSTVq,(FVe*eJ;6kdPZ>M1&O)\,edAm?P1$7r@gK<<+s#*#7lCf&<4CR[St/3)ud,r\L:l]TrAJk8rtn#sgkZ1_LWk.43V>l(l*oJ3%48)nkU*R:nYr>=e=$HQ"Sc]g7[8.D''D'^`tXfIOa<p?pR@AG$25q]OD`3M:e^,B0Uh!%kiRRs1lh]%#dIKPYZ46Z~>endstream
+endobj
+14 0 obj
+<<
+/Filter [ /ASCII85Decode /FlateDecode ] /Length 852
+>>
+stream
+Gat=hhf%4&&BE](/+*QIVW,Sj<G8dqNRA98/R?H8*dBDr!EW$dfd7a>rQ]+J]*^9lMLpG.^:Wjsc:>^gLGlnM!Uq[pcaOUe&*X:384#l9@$^M8pLn>f3JMbkUG8n`lHr0sJbVOl:I6145_I5[oX9O8B]g>J($O6S)E3,ccKelHfnbSuLEWHhOFD4YIcBs;W3?a(`ZH64)NV8IY4:g<^L$Nq/I\M>$/C<d?r;!iUkIi9f:lkt28!\`2(G(o".5E_G9$$`ao#cMojJPZ'W`'D?GPQ/\k(tu+QnG.NRc8D*#KO9X=tF=Qa4ea@)TJf:/bjm'U;]V0=^m36+(Eo2@Z#+E]&O(3+$7BF:N$f):KS7$D%W+_I<h#817kZ]/+-T7A9g>gZ<5GOaos]nMO"75"t-@pb)tT^)q]4I!H1O(QWg4^D`9ATX!Q*5eBqm+1.(h-eq[WLdQcQ:/OauZ!'B^b_hjA5+4R/N^f0SNb;rHoL[EMQ(scj,1NRu-Y//_U?j;<U)aW[F9.dZ.Kl,k<S2$(mlU*c-ppm98S<,\6A4f_+U"2(UK1SuAqVX(\!lopD0Q"Xjm(*YNB?a/7ilT8/JUm7hc@D=s2o"K-4UVo=R,7,h0!,0_]lb5D%AL.mlTr"P;)noP7TV,lnquT.I?"-KP"<"p&h25[;-JoFj9M.9VLqGI+ciV2Gc))>(!U%a&<ukB&2g"&URh"ofauY6<VM1HE5T118e(l2h@5f*h4^d;pUgbG+ughiMH\5mPQZn7Ta&d<lSMpn<'\pL,4MGY>aO"#eN*\qco@iGe=pM;;cr([W+S'RcCoH9c)]ViNtN[(ct6S6>:WK1oh,TJO&ZWCt5G%~>endstream
+endobj
+15 0 obj
+<<
+/Filter [ /ASCII85Decode /FlateDecode ] /Length 677
+>>
+stream
+Gat=(9okbt&A@7.bU!n$8r='$cs)uWX-NBGgN)M/M[P9L_KKZUM>#hN$m]]$fi3h[+_Bg3c+KAJ"<ot(Io[3?@fc$^'-J!H#"ULFTX]WD@e+;$7cZ=W1f"^B[5j(7Se@F>GX:p(k"/8+AIPH,+;$VS,F_)-$7,"a1X?te@h!Y.1FR]=n\u)b_K.adNJ9`KZ/K8D?fH7#V.YcMCXaHT6EoaBOH;dHXL+3q_1rt6cJJ;ZH?DA7J7?bhC?qN2@'&4)@lYCO.OVU/a=pqXa%O=sTVQ<"&)Gm3dBoF[Sf8.KF/%K2,`+<.^'2F@qB.^F"1mj"[:%"U2.m2j8RGr,#0"#oW"H,U0A8hClYThbLg)=;MMLpGGB(9Ec*%^+&$U[nI)hMYTW;gC%+^,aQqTe)YX&YEn$mgcn,;\Aq-mINJV:Cs/"'iuB0SEeHf<PkXhu6tY*!X4[*t<.lpV^+3PF:'$sa\cMSnL4rI4eW,N9su$OuS(k68t&]Js)G/VaX[HrrO\^&8Gb8D%'moqjEd.]o5gSaTfa^<!HYStkjP_YRD%D4/`KfluPN)"W%TL)M<)oHaT[YQ$$J74-;PU<3(kYZ_6T>.%eTdF@,:F&(5;#0$=EHN70q.b<8rSYF+E73`!?rnjA?L`Zq^Atut5n&+PT-jd674^7GC&km`VYIno1~>endstream
+endobj
+16 0 obj
+<<
+/Filter [ /ASCII85Decode /FlateDecode ] /Length 597
+>>
+stream
+GatU.bAQ&g&A7ljk#BueTnnpaa#t,-P6MI-3,#;l4m54*CR_:L3d`3the-Z8DB;8DTHGA>gteBhi<lBn`9(:5^m>[(k5muES3u@/Nq*fibMHN#lnZW%-mf*Fo,VJ%4TLhU&BD'6kZgFg&L%gdQkTUYbXQW9,KVa(NAK2jqIdh4DLi<XBj0>,kH4=/F9@c:-pL2Pn%;s^([2lNVMqV#SaX,<aIR3LA3+!Fc#J<7Ind=b>O!,A^^X*tpp;4"W8<d3;(`1K=^K!k0QUJn:TpoSIR6-;1e#$,g3$/VRQl5pQ%S@p44\IKSI(>NoAE^hbWT*>>cPEoaas6bL^Xe^^Kg?4Er";XIoe7*q);K`^GRXt[F1A;q*Wl6Ah-q?\&%m-($/W1$^rq^DCct?h`Er6^LSdo#0SUlh?(CR@,dS\BcCZq_i?>'Z#/7`=#/*.os'G5k&SUN2E:M0ic@3k_%$;G'U=o:84HT+a"^g.hDK:0OoJj!``g[)G"<.(N?tuTbUlOs=)-30IV;(+rsFa/UtFD/2%*&0%`FGrWig('eSI\Dk-qM$IiRm9A@5jjpQ.C=L0MZ+HMIUhpd<E3NBlq=U@B"7~>endstream
+endobj
+17 0 obj
+<<
+/Filter [ /ASCII85Decode /FlateDecode ] /Length 554
+>>
+stream
+Gat=f;/b/B'SYH?($C]Z,E&*LKh']L%;?O!,&OM1Z%LfLB!tj]^Yb#Z<C\D`9hlnu`]E9E^6^">`T"AlJ<)!?Ht$J'V#ni(*9A!\b/',>4.X#%UR2a;Qtcq#fq`"HG]q=Z3KcL7fhD\&$qQ2fBTlUi\f@f-]&R+>38Ntlr@)NHU4(Ie_9-L+r&^.(.'IsK#?lU<([9?CPru&=JfWhDZI(#dD:5q!K-<m^BM91k%In9c%;P:2^%]57o)<gCSq"ee%U3%6g"&(8#QGjfdMXMXCaH`3N)mM&-=HWa)el+r!O_9qn\L&)lehV-D6TD0kBHO^r?t;VKW1_9-]D`7fX)nOPRIfD4Xf./jC7cZYL2sbana]Pe+Qoa4F'OP.>:`LSE^[U:->oi!l6dXE#id<NScf(-+(QZ[),fBJE?e5S%XM]918uUDcaZnk\j8agGZoM:2g!SBqqcq7*sF-g7bh.Y*D;e%W`6URU1S/YNGa>j9:Riq,HOtWFd!fU#p?nB69&4iFTH!cIVkB?m34]Y58@NeRHHq?B!<sSTA6/$`>3<'*~>endstream
+endobj
+xref
+0 18
+0000000000 65535 f 
+0000000061 00000 n 
+0000000112 00000 n 
+0000000219 00000 n 
+0000000331 00000 n 
+0000000526 00000 n 
+0000000721 00000 n 
+0000000798 00000 n 
+0000000993 00000 n 
+0000001188 00000 n 
+0000001383 00000 n 
+0000001453 00000 n 
+0000001734 00000 n 
+0000001818 00000 n 
+0000002584 00000 n 
+0000003527 00000 n 
+0000004295 00000 n 
+0000004983 00000 n 
+trailer
+<<
+/ID 
+[<c41aa1ae0e9b9080f699bc9a3fa57e7c><c41aa1ae0e9b9080f699bc9a3fa57e7c>]
+% ReportLab generated PDF document -- digest (opensource)
+
+/Info 11 0 R
+/Root 10 0 R
+/Size 18
+>>
+startxref
+5628
+%%EOF

open_pharma_plugins_field_training/models.py

@@ -0,0 +1,265 @@
+"""Pydantic output models for field-training learning packages, assessments, and role-play.
+
+Every generated content piece traces back to a specific source document and
+page/slide via SourceReference, ensuring compliance with approved messaging.
+"""
+
+from __future__ import annotations
+
+from pydantic import BaseModel, ConfigDict, Field, model_validator
+
+
+class StrictTrainingModel(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+
+# ---------------------------------------------------------------------------
+# Document content store
+# ---------------------------------------------------------------------------
+
+
+class DocumentPage(StrictTrainingModel):
+    """A single page (PDF) or slide (PPTX) of extracted content."""
+
+    page_number: int = Field(description="1-indexed page or slide number")
+    text: str = Field(description="Extracted body text from the page or slide")
+    page_type: str = Field(description="Content type: 'page' (PDF) or 'slide' (PPTX)")
+    slide_title: str | None = Field(default=None, description="Slide title shape text (PPTX only)")
+    speaker_notes: str | None = Field(default=None, description="Presenter notes attached to the slide (PPTX only)")
+
+
+class IngestedDocument(StrictTrainingModel):
+    """A fully ingested document stored in the content index."""
+
+    document_id: str = Field(description="Stable identifier derived from the file path")
+    file_name: str = Field(description="Original file name")
+    file_path: str = Field(description="Absolute path to the source file")
+    file_type: str = Field(description="File format: 'pdf' or 'pptx'")
+    title: str | None = Field(default=None, description="Extracted document title")
+    total_pages: int = Field(description="Total number of pages or slides")
+    pages: list[DocumentPage] = Field(description="Page-level extracted content")
+    ingested_at: str = Field(description="ISO-8601 timestamp of ingestion")
+
+
+# ---------------------------------------------------------------------------
+# Source reference (shared across all outputs)
+# ---------------------------------------------------------------------------
+
+
+class SourceReference(StrictTrainingModel):
+    """A traceable reference to a specific passage in an approved source document."""
+
+    document_id: str = Field(description="ID of the source document in the content store")
+    document_name: str = Field(description="File name of the source document")
+    page_number: int = Field(description="Page or slide number where the content appears")
+    excerpt: str = Field(description="Relevant passage from the source supporting the claim")
+
+
+# ---------------------------------------------------------------------------
+# Learning package
+# ---------------------------------------------------------------------------
+
+
+class KeyMessage(StrictTrainingModel):
+    """An approved key message extracted from source documents."""
+
+    message: str = Field(description="The key message or approved claim")
+    category: str = Field(description="Message category: efficacy, safety, dosing, MOA, differentiation, etc.")
+    sources: list[SourceReference] = Field(
+        min_length=1,
+        description="Source references supporting this message",
+    )
+
+
+class LearningObjective(StrictTrainingModel):
+    """A learning objective for a training module."""
+
+    objective: str = Field(description="What the learner should be able to do after training")
+    bloom_level: str = Field(description="Bloom's taxonomy level: 'remember', 'understand', or 'apply'")
+
+
+class TalkingPoint(StrictTrainingModel):
+    """A situation-response pair for field rep conversations."""
+
+    situation: str = Field(description="The conversational context, e.g. 'When an HCP asks about efficacy...'")
+    approved_response: str = Field(description="The approved response the rep should deliver")
+    supporting_data: str | None = Field(default=None, description="Key data points that support the response")
+    sources: list[SourceReference] = Field(
+        min_length=1,
+        description="Source references grounding the approved response",
+    )
+
+
+class Objection(StrictTrainingModel):
+    """A common HCP objection with an approved response."""
+
+    objection: str = Field(description="The objection an HCP might raise, e.g. 'What about the cardiac safety signal?'")
+    approved_response: str = Field(description="The approved response addressing the objection")
+    sources: list[SourceReference] = Field(
+        min_length=1,
+        description="Source references grounding the response",
+    )
+
+
+class LearningModule(StrictTrainingModel):
+    """A single training module within a learning package."""
+
+    title: str = Field(description="Module title")
+    product: str | None = Field(default=None, description="Product name if product-specific")
+    therapeutic_area: str | None = Field(default=None, description="Therapeutic area covered by the module")
+    objectives: list[LearningObjective] = Field(description="Learning objectives for this module")
+    key_messages: list[KeyMessage] = Field(description="Approved key messages the rep must know")
+    talking_points: list[TalkingPoint] = Field(description="Situation-response talking points for HCP conversations")
+    common_objections: list[Objection] = Field(description="Common HCP objections with approved responses")
+
+
+class LearningPackage(StrictTrainingModel):
+    """A complete source-grounded learning package for field reps."""
+
+    title: str = Field(description="Package title")
+    modules: list[LearningModule] = Field(description="Training modules")
+    source_documents: list[str] = Field(description="Names of all source documents used to build the package")
+    generated_at: str = Field(description="ISO-8601 timestamp of generation")
+
+
+# ---------------------------------------------------------------------------
+# Assessment
+# ---------------------------------------------------------------------------
+
+
+class MCQOption(StrictTrainingModel):
+    """A single option in a multiple-choice question."""
+
+    label: str = Field(description="Option label: A, B, C, or D")
+    text: str = Field(description="Option text")
+
+
+class MCQuestion(StrictTrainingModel):
+    """A multiple-choice knowledge check question."""
+
+    question_id: str = Field(description="Unique question identifier")
+    question: str = Field(description="The question text")
+    options: list[MCQOption] = Field(description="Answer options (typically 4)")
+    correct_answer: str = Field(description="Label of the correct option (e.g. 'B')")
+    explanation: str = Field(description="Explanation of why the correct answer is right, citing the source")
+    source: SourceReference = Field(description="Source reference for the correct answer")
+    difficulty: str = Field(description="Difficulty level: 'easy', 'medium', or 'hard'")
+
+
+class ScenarioQuestion(StrictTrainingModel):
+    """A scenario-based question simulating an HCP interaction."""
+
+    question_id: str = Field(description="Unique question identifier")
+    scenario: str = Field(description="The scenario description, e.g. 'A cardiologist asks you about...'")
+    hcp_persona: str = Field(description="HCP persona details, e.g. 'Skeptical oncologist, concerned about toxicity'")
+    ideal_response_points: list[str] = Field(description="Key points that an ideal response should cover")
+    sources: list[SourceReference] = Field(
+        min_length=1,
+        description="Source references for the ideal response points",
+    )
+    difficulty: str = Field(description="Difficulty level: 'easy', 'medium', or 'hard'")
+
+
+class Assessment(StrictTrainingModel):
+    """A complete assessment with MCQ and scenario-based questions."""
+
+    title: str = Field(description="Assessment title")
+    mcq_questions: list[MCQuestion] = Field(description="Multiple-choice knowledge check questions")
+    scenario_questions: list[ScenarioQuestion] = Field(description="Scenario-based application questions")
+    total_points: int = Field(description="Total points available")
+    passing_score_pct: float = Field(description="Passing threshold as a fraction, e.g. 0.8 for 80%")
+    source_documents: list[str] = Field(description="Names of source documents the assessment is based on")
+    generated_at: str = Field(description="ISO-8601 timestamp of generation")
+
+
+# ---------------------------------------------------------------------------
+# Role-play
+# ---------------------------------------------------------------------------
+
+
+class RoleplayTurn(StrictTrainingModel):
+    """A single turn in an interactive role-play conversation."""
+
+    speaker: str = Field(description="Who spoke: 'hcp' or 'rep'")
+    message: str = Field(description="What was said in this turn")
+
+
+class ClaimEvaluation(StrictTrainingModel):
+    """Evaluation of a single claim made during role-play."""
+
+    claim: str = Field(description="The claim or statement being evaluated")
+    status: str = Field(description="Evaluation status: 'correct', 'missed', 'unsupported', or 'inaccurate'")
+    source: SourceReference | None = Field(
+        default=None,
+        description="Source reference if the claim is correct or has a known approved version",
+    )
+    feedback: str = Field(description="Specific feedback explaining the evaluation")
+
+
+class RoleplayScorecard(StrictTrainingModel):
+    """Post-session scorecard summarizing role-play performance."""
+
+    hcp_persona: str = Field(description="The HCP persona used in the role-play")
+    topic: str = Field(description="Product or topic discussed")
+    turns: list[RoleplayTurn] = Field(description="Full conversation transcript")
+    claims_evaluated: list[ClaimEvaluation] = Field(description="Evaluation of each claim made by the rep")
+    score: float = Field(ge=0.0, le=1.0, description="Overall score from 0.0 to 1.0")
+    strengths: list[str] = Field(description="What the rep did well")
+    areas_for_improvement: list[str] = Field(description="Specific areas where the rep can improve")
+    source_documents: list[str] = Field(description="Source documents used for evaluation")
+
+
+# ---------------------------------------------------------------------------
+# Pre-session role-play kit
+# ---------------------------------------------------------------------------
+
+
+class FacilitatorPrompt(StrictTrainingModel):
+    """A facilitator cue for one stage of a role-play session."""
+
+    stage: str = Field(description="Session stage, such as opening, probe, objection, or close")
+    prompt: str = Field(description="Prompt the facilitator delivers to the representative")
+    coaching_intent: str = Field(description="Behavior or skill the facilitator should observe")
+
+
+class EvaluationCriterion(StrictTrainingModel):
+    """One weighted criterion in the pre-session evaluation rubric."""
+
+    criterion: str = Field(description="Capability being evaluated")
+    weight_pct: int = Field(ge=1, le=100, description="Percentage weight for this criterion")
+    evidence_to_observe: list[str] = Field(description="Observable behaviors that demonstrate the criterion")
+
+
+class RoleplayKit(StrictTrainingModel):
+    """A source-grounded kit used to facilitate and assess a future role-play."""
+
+    title: str = Field(description="Role-play kit title")
+    topic: str = Field(description="Product or approved-message topic")
+    hcp_persona: str = Field(description="HCP persona used in the scenario")
+    scenario: str = Field(description="Situation the representative will practice")
+    objectives: list[LearningObjective] = Field(min_length=1, description="Session learning objectives")
+    key_messages: list[KeyMessage] = Field(
+        min_length=1,
+        description="Approved messages the representative should communicate",
+    )
+    common_objections: list[Objection] = Field(
+        min_length=1,
+        description="Likely objections with approved responses",
+    )
+    facilitator_prompts: list[FacilitatorPrompt] = Field(
+        min_length=1,
+        description="Prompts that guide the practice conversation",
+    )
+    evaluation_rubric: list[EvaluationCriterion] = Field(
+        min_length=1,
+        description="Weighted evaluation criteria",
+    )
+    source_documents: list[str] = Field(min_length=1, description="Names of all approved documents used")
+    generated_at: str = Field(description="ISO-8601 timestamp of generation")
+
+    @model_validator(mode="after")
+    def rubric_totals_100_percent(self) -> RoleplayKit:
+        total = sum(item.weight_pct for item in self.evaluation_rubric)
+        if total != 100:
+            raise ValueError(f"evaluation_rubric weight_pct values must total 100, got {total}")
+        return self

open_pharma_plugins_field_training/tools/get_document_page.py

@@ -0,0 +1,67 @@
+"""get_document_page — retrieve the full text of a specific page or slide."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+
+class GetDocumentPageArgs(BaseModel):
+    document_id: str = Field(description="Document ID (from list_documents or search_content results)")
+    page_number: int = Field(ge=1, description="1-indexed page or slide number")
+
+
+TOOL: dict[str, Any] = {
+    "name": "get_document_page",
+    "description": (
+        "Get the full text of a specific page (PDF) or slide (PPTX) from an "
+        "ingested document. Returns the complete text content, slide title, and "
+        "speaker notes (for PPTX). Use this for precise citation when building "
+        "learning content or verifying a claim found via search_content."
+    ),
+    "args": GetDocumentPageArgs,
+}
+
+
+def handle(arguments: dict[str, Any]) -> list[dict[str, Any]]:
+    import json
+
+    from .._content_store import load_document
+
+    document_id = arguments["document_id"]
+    page_number = arguments["page_number"]
+
+    doc = load_document(document_id)
+    if doc is None:
+        return [{"type": "text", "text": json.dumps({"error": f"Document '{document_id}' not found"})}]
+
+    page = None
+    for p in doc.get("pages", []):
+        if p["page_number"] == page_number:
+            page = p
+            break
+
+    if page is None:
+        return [
+            {
+                "type": "text",
+                "text": json.dumps(
+                    {
+                        "error": f"Page {page_number} not found in document '{document_id}' (total pages: {doc.get('total_pages', 0)})"
+                    }
+                ),
+            }
+        ]
+
+    output = {
+        "document_id": document_id,
+        "document_name": doc["file_name"],
+        "document_title": doc.get("title", ""),
+        "page_number": page["page_number"],
+        "page_type": page["page_type"],
+        "text": page["text"],
+        "slide_title": page.get("slide_title"),
+        "speaker_notes": page.get("speaker_notes"),
+    }
+    return [{"type": "text", "text": json.dumps(output, indent=2)}]

open_pharma_plugins_field_training/tools/ingest_document.py

@@ -0,0 +1,147 @@
+"""ingest_document — extract structured text from a PDF or PPTX file."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from pydantic import BaseModel, Field
+
+
+class IngestDocumentArgs(BaseModel):
+    file_path: str = Field(description="Absolute path to the PDF or PPTX file to ingest")
+    document_title: str | None = Field(
+        default=None,
+        description="Optional title override. If not provided, the title is extracted from the first page/slide.",
+    )
+
+
+TOOL: dict[str, Any] = {
+    "name": "ingest_document",
+    "description": (
+        "Extract structured text from a PDF or PPTX file and store it in the "
+        "training content index. Returns a summary with the document ID, title, "
+        "page count, and file type. For PPTX files, extracts slide titles, body "
+        "text, and speaker notes per slide. For PDF files, extracts the embedded "
+        "text layer per page. Use this before search_content or get_document_page."
+    ),
+    "args": IngestDocumentArgs,
+}
+
+
+def handle(arguments: dict[str, Any]) -> list[dict[str, Any]]:
+    import json
+    import os
+    from datetime import datetime, timezone
+
+    from .._content_store import document_id_from_path, save_document
+
+    file_path = arguments["file_path"]
+    document_title = arguments.get("document_title")
+
+    if not os.path.isfile(file_path):
+        return [{"type": "text", "text": json.dumps({"error": f"File not found: {file_path}"})}]
+
+    ext = os.path.splitext(file_path)[1].lower()
+
+    if ext == ".pdf":
+        pages, auto_title = _extract_pdf(file_path)
+        file_type = "pdf"
+    elif ext in (".pptx", ".ppt"):
+        pages, auto_title = _extract_pptx(file_path)
+        file_type = "pptx"
+    else:
+        return [{"type": "text", "text": json.dumps({"error": f"Unsupported file type: {ext}. Use .pdf or .pptx"})}]
+
+    pages = [p for p in pages if p.get("text", "").strip()]
+
+    title = document_title or auto_title or os.path.basename(file_path)
+    doc_id = document_id_from_path(file_path)
+
+    doc = {
+        "document_id": doc_id,
+        "file_name": os.path.basename(file_path),
+        "file_path": os.path.abspath(file_path),
+        "file_type": file_type,
+        "title": title,
+        "total_pages": len(pages),
+        "pages": pages,
+        "ingested_at": datetime.now(timezone.utc).isoformat(),
+    }
+
+    save_document(doc)
+
+    summary = {
+        "document_id": doc_id,
+        "file_name": doc["file_name"],
+        "file_type": file_type,
+        "title": title,
+        "total_pages": len(pages),
+        "ingested_at": doc["ingested_at"],
+    }
+    return [{"type": "text", "text": json.dumps(summary, indent=2)}]
+
+
+def _extract_pdf(file_path: str) -> tuple[list[dict], str | None]:
+    import pypdfium2 as pdfium
+
+    doc = pdfium.PdfDocument(file_path)
+    pages: list[dict] = []
+    auto_title: str | None = None
+
+    for i in range(len(doc)):
+        textpage = doc[i].get_textpage()
+        text = textpage.get_text_bounded()
+        textpage.close()
+        pages.append(
+            {
+                "page_number": i + 1,
+                "text": text,
+                "page_type": "page",
+                "slide_title": None,
+                "speaker_notes": None,
+            }
+        )
+        if i == 0 and text.strip():
+            first_line = text.strip().split("\n")[0].strip()
+            if first_line and len(first_line) < 200:
+                auto_title = first_line
+
+    doc.close()
+    return pages, auto_title
+
+
+def _extract_pptx(file_path: str) -> tuple[list[dict], str | None]:
+    from pptx import Presentation
+
+    prs = Presentation(file_path)
+    pages: list[dict] = []
+    auto_title: str | None = None
+
+    for i, slide in enumerate(prs.slides):
+        title = ""
+        if slide.shapes.title:
+            title = slide.shapes.title.text
+
+        body_texts: list[str] = []
+        for shape in slide.shapes:
+            if shape.has_text_frame:
+                body_texts.append(shape.text_frame.text)
+
+        notes = ""
+        if slide.has_notes_slide and slide.notes_slide.notes_text_frame:
+            notes = slide.notes_slide.notes_text_frame.text
+
+        pages.append(
+            {
+                "page_number": i + 1,
+                "text": "\n".join(body_texts),
+                "page_type": "slide",
+                "slide_title": title or None,
+                "speaker_notes": notes or None,
+            }
+        )
+
+        if i == 0 and title:
+            auto_title = title
+
+    return pages, auto_title